A telemetry technique has been developed to locate a mobile person such as a wandering dementia patient. In the field experiments, some problems were discovered that had to be solved for the practical use of this technique. These included: the effect of weather conditions, the problem of network interruptions of a mobile phone, the problem of the locations where the GPS radio signal could not be recovered, and the problem of the location error of GPS. The solutions for these problems were worked out, making the proposed technique more practical.

The problem of wandering dementia patients has become an increasingly important issue in our society (Negley et al. 1990, Yamamoto and Wakamatsu 1991, Hosaka et al. 1996). We have developed a technique to locate a wandering person using a mobile phone and the global positioning system (GPS). An experimental system was manufactured, and the feasibility of this technique has been studied in fundamental experiments (Shimizu et al. 1998). Based on these experiments, some problems were found that need to be solved for practical application of this technique. This paper discusses the problems and presents some possible solutions.

Figure 1 illustrates a block diagram of our basic approach. The location system consists of a personal computer at a home (or institution) and a responder carried by the patient. The responder has two functions. One is to receive the location signals from GPS satellites. Another is to send the location back to the home.

Figure 1. Location system for a wandering person.

When a helper (usually a family member or nurse) realizes the absence of a patient, the helper calls the responder using a mobile telephone network. In the responder, the location of the patient is calculated using signals from GPS satellites. The information on the location is automatically sent back from the responder to a personal computer through the mobile telephone network. In the personal computer, the information is processed and displayed on a CRT display. The helper finds the position of the patient on the CRT display and makes arrangements to protect the patient.

In studies using the experimental system, the following problems were found.

The frequency of successful locations can be dependent on weather and terrain conditions.

We are unable to obtain location information when the wandering person enters areas such as underground streets or tunnels where the GPS radio signal cannot reach.

In order to reduce battery consumption, the responder was initially designed to automatically turn off the power supply for the location function when the communication link of the mobile telephone was disconnected. Occasionally the connection of the mobile telephone line was interrupted momentarily. This interruption resulted in the restart of the GPS location algorithm, which required several minutes before establishing the locating operation. This problem became serious when the connection of the mobile telephone was not stable.

The accuracy of location with our system is usually within 100-200 m. In most cases, this is sufficient to find a wandering person. However, it would be better if we can improve the accuracy of location.

To solve these problems, a test system was manufactured (Shimizu et al. 1998) and the following study was conducted.

The communication link of the mobile telephone is known to be fairly invulnerable to weather and terrain conditions. However, the GPS radio wave can be affected by these conditions particularly when the responder is carried in close contact to the subject’s body. To examine the effect of these conditions, the performance of our system was evaluated under adverse conditions.

The responder was attached to the body of a subject with the GPS receiver kept vertical in close contact with his chest. The subject walked along the streets in different areas, such as a university campus and a city area with tall buildings. The positions of the subject were estimated using the GPS location data every second. Figure 2(a) shows an example of the results. The subject walked along a street from Point A to Point B. The positions recorded by the base station are shown in a series of dots on the same map. The separated points correspond to the position when the location could not be obtained. The location was successfully determined in most places and the precision was satisfactory for practical use.

There were some points where the location data were not available probably due to the shadowing effect of tall buildings. However, the interruptions in locating the subject were only several minutes at the most. This period is sufficiently short to find the wandering person in practice.

Figure 2(b) shows the distribution of the location error in a horizontal plane. Even with the effect of tall buildings, the accuracy was within 100 m in all directions. This was satisfactory for practical use, as well.

Next, the effect of bad weather was investigated. The GPS was not affected by heavy rain. However, falling snow is charged with static electricity and induces noise when passing antenna elements. The effect of falling snow was investigated in this study. A location experiment similar to the one described was conducted in a heavy snowstorm. Figure 3 shows the result. The number of successful locations decreased a little, but they were still frequent enough for the purpose of finding the wandering person in practice. The error was within 150 m, which was satisfactory, as well.

Figure 2. Result of determining location in different areas (from university campus to city): (a) trace of the travel route and obtained GPS locations, (b) horizontal distribution of location error.

Figure 3. Results of determining location in a heavy snowfall.

When a wandering person enters an area where GPS radio signals cannot reach, it is not possible to determine location. The mobile telephone link is usually accessible in these areas; thus, we modified the system so that the responder can obtain the position of the wandering person periodically and memorize that position until the next position is determined. In this way, when the base station attendant calls the responder, it can return the position of the wandering person just before entering the blind spot. This information can provide a significant clue to locating the person.

The problem of intermittent disconnection can be solved using the same approach. Figure 4 shows a timing diagram of the automatic intermittent activation of the system. When the mobile telephone link is in a waiting mode, the function of the GPS location finder is activated 6 minutes in every 21 minutes. The location data obtained in the period of the GPS operation is stored in the memory of the responder. As long as the GPS signal is available, the location data is renewed every 21 minutes. If the GPS locations become impossible to determine, the most recent data is kept in the memory.

Figure 4. Timing diagram of automatic intermittent activation to address the problem of blind spots and momentary disconnection.

When the base station calls the responder and the mobile line is connected, the intermittent operation of the GPS is shifted to the continuous operation mode. When the mobile telephone link is disconnected, the intermittent operation is resumed from the activated (ON) mode of the GPS. Therefore, the function of the GPS is kept ON at least 6 minutes after the mobile telephone link is disconnected. In this way, even if the telephone link is disconnected momentarily, the host computer reconnects the link by automatically redialing and the base station attendant can obtain the location data without the delay of initializing the GPS operation.

To make this system more useful, the location error should be minimized. We have applied the differential GPS (DGPS) technique to improve location accuracy. Figure 5 shows the result of an experiment to show the improvement with the DGPS technique. In the experiment, the responder was placed in a fixed position and the location estimated by GPS was recorded for about 10 minutes. Using the differential technique, the location error was decreased significantly.

Figure 5. Improvement in accuracy of location by DGPS: (a) before the correction, (b) after the correction by DGPS.

Many problems were found and solutions were worked out for the practical application of the location system for wandering dementia patients. In a strong snowstorm, the number of estimated locations decreased, but this decrease was not so serious as to hinder the search for the wandering person. We could obtain his location within a satisfactory error range and with a practical frequency. To solve the problem of the network interruption and the unavailability of the GPS radio signal, a countermeasure was devised. The mobile terminal was made to obtain the location of the moving person periodically and to store the last location in memory. When it was called up from the fixed station, it returned the most recently determined location. The accuracy of the location estimation was improved by the use of the differential GPS principle.

The usefulness of the developed system was enhanced significantly as our experience with it increased. The last problem remaining is the miniaturization of the responder. Today, a card-type or a watch-type GPS receiver is commercially available. The size of the mobile telephone terminal also can be reduced by limiting its function to only data transmission. If a small responder is developed, it will be useful not only to detect wandering dementia patients, but also to address the general need to locate a person in various other applications.

A part of this research was supported by a grant from the Ministry of Education, Science, Sports and Culture, Japan.

Pill-shaped biotelemeters originally designed for space flight applications will soon be used for monitoring the health of a fetus during and after in-utero fetal surgery. The authors have developed a family of biotelemeters that are not only small enough for rodent studies on board the space shuttle or international space station, but also fit through a 10 mm trocar, a plastic tube that is used in endoscopic fetal surgery to obtain minimally invasive access to the fetus. The first "pill" measures pressure and temperature, and is currently undergoing long-term leakage and biocompatibility tests. A second pill under development measures pH and temperature. A prototype of the "pH-pill" has been built and successfully tested and is presently being miniaturized into the same dimensions as the "pressure pill." Additional pills measuring heart rate, ECG, other ions such as calcium and potassium, and eventually glucose and blood gases, will follow. All pills are designed for ultra-low power consumption yielding lifetimes of up to 10 months in order to meet the requirements of fetal monitoring, and to provide the capability for long-term space station experiments. Each pill transmits its pulse-interval-modulated signal on a unique carrier frequency in the frequency range of 174-216 MHz. A custom-designed multi-channel receiver demodulates and decodes each pill’s signal and sends the data to a LabVIEW program that performs real-time data analysis and display.

Through its Advanced Technology Development - Biosensors Program (ATD-B), the National Aeronautics and Space Administration (NASA) Sensors 2000! (S2K!) program seeks to develop advanced technologies to monitor physiological parameters during space flights. Research and development efforts are focused on designing, building, and demonstrating telemetry-based sensing systems that measure physical, chemical and biological parameters (Hines 1996). Not only will these systems provide valuable data for studying the effects of microgravity on biological subjects in space, but they will also be of great benefit for monitoring patients on earth. We focus here on NASA sensor systems being adapted to improve the ability of doctors to monitor the human fetus and mother following in utero fetal surgery (Hines et al. 1995).

Pediatric surgeons at the University of California San Francisco Fetal Treatment Center (FTC) have developed a revolutionary surgical procedure to treat fetuses suffering from diaphragmatic hernia, a condition in which a hole in the diaphragm allows internal organs to shift from the abdominal cavity into the chest cavity (Harrison et al. 1997). The lungs have insufficient space to develop and about 60 percent of children born with this condition die. The FTC doctors first used an open hysterotomy (similar to a Cesarean section) surgery to access the fetus and correct this anomaly. More recently they developed a minimally invasive procedure using endoscopic techniques they refer to as FETENDO (Figure 1). With this approach, in utero surgical procedures are conducted through a number of small openings in the uterus, the largest being 10 mm. This approach was developed to minimize postoperative preterm labor, a problem that is encountered in all patients who undergo fetal surgery. Accurate monitoring of uterine contractions in the postoperative period is critical to develop medications that can inhibit the progression of preterm labor.

Figure 1. Endoscopic Fetal Surgery.

In 1993, the FTC established a collaborative development effort with S2K! to adapt NASA’s implantable biotelemetry devices to the monitoring needs of the human fetus and its uterine environment. One of their major goals was to accurately measure small intra-uterine pressure changes in order to diagnose and treat preterm labor. While the first FTC system used was based on a commercial sensing device, NASA’s current efforts involve the design of pill-sized transmitters that are small enough to be introduced into the uterus through a 10 mm trocar. A pill-transmitter measuring pressure and temperature (Figure 2), as well as a receiver and data acquisition system to record the transmitted signal, has been designed and built.

Figure 2. Pressure/Temperature Pill-Transmitter.

The transmitter is capable of measuring changes in pressure and temperature in utero over several months. It is the first of a family of pill transmitters that will provide NASA and the pediatric surgeons with otherwise unattainable information on intra-uterine pressure and temperature as well as fetal heart rate and tissue pH, parameters critical to the successful monitoring and evaluation of postoperative preterm labor and fetal health and well being.

The current biotelemetry system consists of four major building blocks: pressure/temperature transmitter, biotelemetry receiver, data acquisition card, and digital data acquisition system. The transmitter uses Pulse Interval Modulation (PIM) to send temperature and pressure information out of the uterine environment. The RF carrier frequency is in the biomedical range (174-214 MHz). A pair of pulses (RF bursts) is transmitted at a frequency of about 1-5 Hz. The interval between successive pulse pairs is proportional to the measured temperature. The interval between the two pulses of a pair is proportional to the sensed pressure. The low data rate and PIM encoding scheme conserves power and guarantees a transmitter lifetime of several months. The data rate is sufficient for monitoring intra-uterine contractions, which occur over several minutes. The transmission range depends on the position of the transmitter in the body, the biological environment, and the antenna orientation; it typically ranges between 3 feet and 6 feet.

The pill-transmitter is small enough to fit through a 10 mm trocar. We used Chip-on-Board technologies (COB) to minimize the size of the printed circuit board (22 mm x 8 mm) (Lau 1994). Unpackaged dies are flip-chip bonded directly onto the printed circuit board, along with surface mount resistors and capacitors. A pressure transducer die is wire bonded to the printed circuit board (Figure 3). Two silver-oxide batteries and the printed circuit board are placed into a pill-shaped shell, which is then encapsulated in a thin layer of conformal coating that acts as a moisture barrier and medical-grade silicone for biocompatibility. The pill-transmitter is 35 mm long and has a diameter of 9 mm. Temperature is sensed by a thermistor, and pressure changes by a solid-state, piezoresistive pressure transducer. Absolute pressure readings are not required; only changes in pressure (typically 40-60 mm Hg) are of interest in monitoring preterm labor. This simplifies the design of the transmitter.

The receiver is a stand-alone unit that converts the pulse interval modulated RF signal into a digital pulse stream, which is then decoded into voltages proportional to temperature and pressure. The RF portion consists of a Konigsberg receiver module TR8 that has been modified to demodulate low-frequency PIM signals. The digital output signal of the TR8 is then processed by the decoder. The decoder extracts and integrates the two pulse intervals that contain information on pressure and temperature. Sample-and-hold amplifiers convert the integrator outputs into analog voltages. The sensitivity of the temperature channel is 10 mV/° C, and of the pressure channel 1 mV/mm Hg.

The data acquisition system is based on LabVIEW (a graphical programming language for virtual instrumentation). The current system uses a PCMCIA card to read the pressure and temperature data from the biotelemetry receiver into a laptop computer where it is processed by the LabVIEW software. The program displays, analyzes, and stores the data as a function of time. It performs peak detection (contractions), integrates the area under the pressure-time curve, and determines the frequency of contractions. Furthermore, it enables the doctor to view the frequency spectrum of the pressure-time curve and a histogram of the intervals between contractions. This information will help the FTC surgeons to determine and quantify the onset of preterm labor more accurately, which is essential for effective treatment.

Several pressure/temperature pill-transmitters have been successfully tested in body-temperature saline and will soon be implanted in rats to test their performance in vivo. A prototype version of a pH/temperature pill-transmitter has been built and will be used to measure rumen pH of cattle to verify the feasibility of pH-biotelemetry in vivo as well as in rats to study acidosis.

Implantable biotelemetry devices, developed by NASA to measure and transmit physiological parameters from biological subjects such as rats and mice during space flights, have a variety of applications on earth as well. One is fetal surgery, which requires accurate postoperative monitoring of the fetus and its intra-uterine environment. Endoscopic surgical approaches, as performed by the FTC surgeons at UCSF, require small pill-sized transmitters with outside diameters less than 10 mm and lifetimes over several months in order to monitor the fetus after surgery. NASA’s S2K! program has successfully developed the first biotelemetry transmitter that meets these requirements. They demonstrate that NASA’s technology is not limited to applications in space alone, but is also being used to enhance and save human lives on earth.

Future commercial applications of these new pill-transmitters go far beyond fetal surgery. Pill-transmitters small enough to be swallowed are currently in development. Patients with digestive disorders could swallow a pill-transmitter that monitors intestinal acidity or gives data about contractions of intestinal smooth muscle, allowing doctors to better diagnose gastrointestinal diseases. There is also great potential in sports medicine. Pill-transmitters can monitor the performance of athletes as they train. In addition, the military has expressed interest in similar devices to monitor the health and performance of soldiers in hostile environments. Future measurements will include pH, ions of interest (e.g. Ca⁺⁺, Na⁺, K⁺), heart rate, ECG, EEG, EMG, blood gases (O₂, CO₂), and glucose.

KULeuven, ESAT-MICAS, Kard. Mercierlaan 94,
B-3001 Heverlee, Belgium

UKC, Electronic Engineering Laboratories,
Canterbury, CT2 7NT, United Kingdom

This paper describes a system for detecting hip prosthesis loosening, based on a mechanical vibration analysis method. The system consists of a circuit that is embedded in a cavity inside the prosthesis. It uses two wireless links. The circuit detects the (induced) vibration and converts it to an electronic signal. One of the links is used to power the implantable circuit; the second one transmits the electronic signal from the implant to a PC.

In the western world, total hip replacement is carried out on 7-8 people out of 10,000 (Lawes 1993). A common problem with hip implants is that they can loosen and cause discomfort, pain, and difficulty in walking. However, various disorders such as infection can have similar symptoms, making diagnosis hard. The usual diagnostic methods to assess loosening are X-ray imaging and nuclear medicine (isotopes). Neither of these methods can give a clear picture about the mechanical interface between the prosthesis and the femur. Hence, a novel technique (vibration analysis) is proposed. Though this method has been adapted in studies before (Rosenstein et al.1989, Collier et al.1993), the novelty presented here lies in the fact that part of the system is implanted: the monitoring part to detect vibration is placed inside the prosthesis head. An overview of the implantable circuit together with external electronics is illustrated in Figure 1. The implantable circuit consists of the accelerometer together with a sensor interface whose signal is digitized by a microcontroller’s on-chip A/D to transmit the data digitally. The system is equipped with a telemetric system in order to communicate with the outside world. The data is received externally and displayed on a personal computer through a data acquisition card.

Figure 1. Overview of the system for detecting hip prosthesis loosening.

Telemetry is the only viable way, since the device is intended for long term (lifetime) implantation. This implicitly means that battery power cannot be considered, and that power needs to be provided by an inductive link.

To detect hip prosthesis loosening, mechanical sine wave vibrations are applied to the femur. An electromagnetic shaker is pushed against the skin near the distal end of the femoral bone. A miniature accelerometer, placed in the prosthesis head, captures the response of the bone-prosthesis system to the applied vibrations. If there is no loosening, the bone-prosthesis system behaves linearly and only the excitation frequency will be detected. However, if there is loosening, the system will behave in a non-linear mode and harmonics will be detected by the accelerometer.

Experiments have shown that an excitation frequency between 100 Hz and 200 Hz is very well suited to detect loosening. The accelerometer bandwidth must be capable of picking up the base frequency and at least two harmonics. In the case of loosening, the first and second harmonics clearly appear in the frequency domain. This distortion is also visible in the time domain. Figures 2a and 2b show the measurement on a loose prosthesis in the time and the frequency domain respectively.

Figure 2. (a) Time domain signals of a loose prosthesis, top waveform shows the applied vibration signal and bottom waveform is the resultant signal pick up by the accelerometer. (b) The same signals in the frequency domain of the vibration signals showing the appearance of higher order harmonics.

The implant is inductively powered, using two coils (Figure 3). The external transmitter coil L₁ is driven by a dedicated Class E amplifier. The receiver tank circuit consists of the coil L₂ and the capacitor C₂. The tank circuit is tuned to the power transmission frequency. The internal power unit consists of a rectifier and a regulator to deliver a stable 5 V supply to the implantable circuit. The entire circuit is placed in a cavity, which measures 1.23 cm³, inside the implant itself.

The power link was designed in two steps. The first step was the design of a coil set. Special attention was given to the power transfer efficiency of the coil link. In a second step, the Class E amplifier was designed. This two-step design will be discussed first. A short description of the internal power unit is given next.

Magnetic design. The receiver coil L₂ is placed inside the hip prosthesis head. The transmitter coil L₁ is placed around the patient’s leg, so the transmitter coil can easily be removed when not required. Although with this configuration only poor coupling can be achieved, it is preferred to the alternative, by which the external coil is placed laterally to the patient’s leg. In that case, misalignment of the two coils can cause mistuning of the driver and loss of efficiency.

Figure 3. Power link and internal power unit.

As the magnetic coupling is very weak (k = 0.48%), it is essential to optimize the link design. This was done, using the exact link formulas (Van Schuylenbergh 1998). The power transfer efficiency of the link is given by:

where , and .

Q_L1 and Q_L2 are the unloaded coil quality factors. R_load is the load of the internal circuit and was chosen to be 800 W , as the power demand of the internal circuit is about 30 mW (including the losses in the rectifier and the regulator), at a supply voltage of 5 V. Optimization of the link efficiency can be achieved by varying the inductance of the secondary coil L₂. As the receiver tank L₂-C₂ is in resonance, a change of inductance L₂ leads to a change of a . The result of this Matlab optimization can be seen in Figure 4. Varying the primary coil inductance has no influence on the efficiency. The maximal theoretical link efficiency h _link of this configuration is 0.43%.

The coil set was designed using inductance formulas provided by Terman (1947). It can be seen in Table 1 that the formulas give a good estimation of the magnetic parameters. Using other formulas (Hochmair 1984) was not as satisfactory due to the large diameter of the primary coil L₁.

Figure 4. Link efficiency optimization by varying L₂.

Table 1. Comparison of calculated and measured values of the magnetic parameters.

Driver design. A Class E topology (Figure 5) was chosen to minimize power losses in the driver (Sokal and Sokal 1975). Ideally, the efficiency of a Class E driver is 100%. In practical designs, an efficiency of 90% can easily be obtained. The main contribution to power loss is the on-resistance of the switching element S. The switch can be either FET or a bipolar transistor. As this driver deals with voltages up to 60 V, a HEXFET power transistor was chosen.

The internal circuit is seen by the transmitter circuit as equivalent impedance Z_eq. As the internal circuit is tuned to the power transmission frequency, this impedance is purely resistive and given by:

This leads to an adapted coil quality factor Q_Lad :

However, due to the small coupling coefficient k, the equivalent resistance R_eq is negligible compared to the coil resistance R_L1 and Q_Lad equals Q_L1.

Figure 5. Class E driver circuit.

The capacitors C_1s and C_1p can be calculated for a given coil, driver frequency and switch duty cycle (Kazimierczuk and Puczko 1987). In this case, a driver frequency of 750 kHz and a switch duty cycle of 50% were chosen. The value of the series capacitor C_1s is mainly determined by the coil inductance L₁, whereas the shunt capacitor C_1p is also determined by the coil quality factor Q_Lad.

Internal power unit. The internal power unit (Figure 1) consists of the receiver L₂-C₂, a Shottky rectifier and a voltage regulator. The coil L₂ picks up a part of the external field produced by L₁. Due to the shunt capacitor C₂, L₂ acts as a voltage source. The power signal is rectified by the Shottky rectifier and stabilized at 5 V by the LP2951 voltage regulator (National Semiconductor). The internal power unit is incorporated in the hip prosthesis head, together with the data collection circuit, which will be discussed next.

A miniature capacitive accelerometer from VTI Hamlin is used together with a matched sensor interface to measure the induced vibrations. The frequency bandwidth of the accelerometer is 2 kHz, which is sufficient to pick up the excitation frequency and two higher harmonics. In comparison to piezoresistive or piezojunction accelerometers, capacitive accelerometers have low power consumption (Puers 1993), which is favorable.

Due to the physical constraints of the implant and because the data only needs to be transmitted through the skin, the transmitter stage is comprised by a simple RLC series resonant circuit. This will produce across the coil L_t, a decaying sine wave for every rising and falling edge of a transmitted pulse, of frequency equal to:

(Figure 6).

The front end of the external receiver circuit is shown in Figure 7. It consists of dual gate MOSFET tuned amplifier. The Capacitors C_r1 and C_r2 are used to tune the amplifier to the transmitted frequency. Inductor L_r1 is the receiver coil, which consists of two turns and is 10 cm in diameter. It is placed against the skin next to the femoral head, to be as close as possible to the implant’s data transmission coil.

The data are acquired on a personal computer, using a DAS-1701AO data acquisition card from Keithley. A software package is being developed under Windows 95 using C++ for acquiring the data from the receiver and displaying it on the computer’s screen. The software is also responsible for controlling the vibration frequency of the electromechanical shaker.

The test set up (Figure 8) is able to transfer 25 mW, which is enough to power the internal circuit (Table 2). The Class E driver power consumption is 2.6 A at 4 V, being 10.4 W. This is an overall efficiency h _total of 0.25%. The power demands of the internal circuit are described in Table 2.

Figure 8. Test setup of the inductive power link.

Table 2. Power demands of the internal circuit.

A next step in the development of this telemetric system is the miniaturization and hermetic packaging of the internal circuit, which should fit into the cavity in the hip prosthesis head. The system will then be ready for clinical tests. Recent cadaver experiments of the system already gave satisfying results.

This research program is supported by the European Community as the BIOMED 2 project No. BMH4-CT97-2173.

Inside the human hip joint, friction between prosthetic head and acetabular cup increases the temperature during activities like walking. Higher implant temperatures may cause tissue damage and contribute to implant loosening. A hollow-shaft hip implant was instrumented to measure local temperatures at ten points along the whole length of the femoral component in vivo. Inside the prosthetic neck, three semiconductor strain gauges sense the deformation for three-dimensional load measurement.

In 1990 hip force measurements in a patient revealed a 3.5 °C temperature increase inside the hip after 45 minutes of normal walking (Bergmann et al. 1991). The temperature sensor was part of a four channel telemetry unit inside the neck of a titanium hip joint primarily used for compensating the thermal influence on the circuit (Graichen and Bergmann 1991). This coincidental observation raised the question of whether repeated friction-induced high implant temperatures may cause tissue damage and contribute to implant loosening. A new hip implant (Figure 1) was therefore instrumented to measure telemetrically forces and temperatures at several points along the entire length of the femoral component.

The hollow shaft hip endoprosthesis is made of titanium alloy. A ceramic ball head acts against a conventional polyethylene or ceramic acetabular cup. The hollow neck contains three semiconductor strain gauges and one of the telemetry units (Figure 2). To measure the neck deformations three semiconductor strain gauges (SG1 to SG3) are mounted around the inner circumference of the neck in an arrangement similar to the previously used, force measuring 4-channel hip implant (Bergmann et al, 1988). The three components of the spatial hip contact force can be calculated from these signals using preoperatively obtained calibration data (Bergmann et al. 1990). Furthermore, two NTC resistors (TS1, TS2) are fixed to the inner wall of the implant neck. Six additional temperature sensors (TS3 to TS8) are arranged inside the hollow shaft.

Figure 1. Cut model and original hip implant.

The shaft contains a coil for the inductive power supply of two telemetry units. Telemetry Unit 1 senses the signals of SG1 to SG3, TS1, TS2, a fixed calibration resistor, the temperature of the hybrid circuit and the rectified supply voltage. Telemetry Unit 2 is arranged inside the shaft above the common power coil. It transmits the signals TS3 to TS8, the hybrid circuit temperature and the supply voltage.

All components are mounted from the top through the hollow implant neck. The temperature sensors TS3 to TS8, the secondary power coil and Telemetry Unit 2 are assembled and completely wired on a complex spring device. It is pressed together and introduced through the neck down to the bottom of the hollow shaft. Once in place, TS3 to TS8 are pressed by the spring against the inner wall of the implant shaft. For permanent fixation, the shaft is filled with silicone rubber. Telemetry Unit 1 is connected with its sensors and the power supply. The two telemetry transmitters are wired to a feedthrough with four niobium leads in the center of the titanium seal plate. The neck of the implant is hermetically closed by electron beam, welding this plate on its top. On the outside, two niobium wires welded to the feedthrough leads form separate one-turn loop antennas. These transmitter coils fit into the small space inside the ceramic head and are thus protected mechanically.

Figure 2. Schematic arrangement of assembled components.

The telemetry device is powered inductively by a magnetic field with a frequency of 4 kHz. It is generated by an induction coil (primary power coil) slipped over the patient’s leg and fixed near the hip joint. An AC current is induced in the secondary power coil L1 inside the implant. Two Zener diodes (D) and the integrated bridge rectifier limit the rectified voltage to +5.1 V (Figure 3). The constant current I0, produced by the current source CS0 is set to 120 µA. The current sources (current mirror circuits) CS1 to CS8 are connected successively to the output of CS0 by an eight channel analog switch. The first six channels are controlled by six sensors, S1 to S6. Channel 7 detects the temperature of the hybrid circuit with the aid of a NTC resistor and channel 8 senses the unregulated supply voltage by a constant integrated resistor. The difference current I_diff therefore depends on the resistance of the sensor just activated by the multiplexer.

Figure 3. Function diagram of the implanted eight-channel telemetry device

The range of the currents Ii of the channels i = 1 to 7 is set by the reference currents Iref i (30 µA), and the relation between the external resistor Rref and the sensor resistances Rsi:

Ii = Iref i * Rref / Rsi with i = 1 to 7

For different applications, the first seven channels are split up into three groups (CH1 to CH4, CH5 to CH6, CH7) with different possible sensor combinations. This permits the use of NTC-thermistors and/or of strain gauges for the first six channels.

While CH1 to CH4 of Telemetry Unit 1 are connected with three semiconductor strain gauges (SG1 to SG3) and one resistor, CH5 and CH6 are used for TS1 and TS2. Telemetry Unit 2 is connected to the six temperature sensors TS3 to TS8 distributed along the prosthetic shaft (Figure 2). The sensitivity of the channels used by strain gauges depends on the strain gauge resistance and the reference resistor Rref SG. Channels with thermistors and CH7 are controlled by Rref NTC (Graichen et al., 1996).

The voltage of capacitor C4, which is charged via the closed switch by the individual difference current Idiff, is sensed by the timer. At a level of +2.4 V, the timer output creates a low level, clocks the 3 bit counter to activate the next channel, and opens the switch, which is held open during the discharge (and transmitting) time of 10µs. When reaching the lower voltage threshold of +1.2 V, the timer output changes to the high level, which closes the switch for the next charge period of C4. After eight cycles, all sensor signals are transformed into time periods generating a pulse train at the timer output. To reach a sampling rate of approximately 250 samples per second for each channel, C4 is chosen to be 10 nF. All components of the customized chip (gray area in Figure 3) are integrated on a 4.2 x 5.0 mm bipolar transistor array. The integrated circuit and the transmitter have a power consumption of about 10 mW. The circuit was developed in cooperation with the Institute of Microelectronics at the Technical University of Berlin and manufactured by the same institute. The output pulses of the integrated timer circuit activate the RF transmitter for a duration of 10 µs. The transmitter consists of a single NPN transistor in a grounded base circuit which is tuned to a frequency in the ranges of 120 MHz or 140 MHz by two different values of L2.

The transistor, the integrated circuit, and the passive discrete components of the telemetry device are mounted on both sides of a 14 mm by 7 mm ceramic substrate (Figure 4). Two layers of gold conductors are printed using thick-film technology on each side and connected via 13 holes filled with silver palladium. Twelve soldering pads for connecting the power coil, transmitter antenna and strain gauges are concentrated on both sides at one end of the substrate. The hybrid is fixed inside a metal cylinder and closed at the top with an encapsulation material (Graichen et al. 1995). The thick-film hybrid circuit was manufactured by HYMEC B.V. Sittard, The Netherlands.

The external telemetry system components are shown in Figure 5. A power generator produces a sine wave voltage of up to 50 Vpp at a frequency range of 3.5 kHz to 4.5 kHz.

The external power coil is series-tuned to the resonant frequency by a capacitor. By phase control of the output voltage and current, the oscillator is automatically tuned to the resonant frequency. Inductance changes, for example caused by metal chairs or coil deformations, are thereby compensated. During the measurements, the coil is slipped over the patient's leg and fixed with a belt (Figure 6). The generator output voltage is controlled by software to a minimum value. This avoids any temperature increase inside the implant caused by magnetic induction. A small loop antenna with an integrated amplifier is placed near the hip joint to receive both radio-frequency (RF) pulse trains of the implant at about 140 MHz and 120 MHz. Twin VHF-receivers demodulate the signals for computation and storage. The 3 force components, 10 temperature values and the induced power are monitored in real time on the LC display of a personal computer. The patient's exercises are filmed by one or two video cameras and recorded together with the implant signals on a S-VHS video cassette recorder with four separate audio tracks. Two high-fidelity tracks record the demodulated pulse trains of the telemetry systems and the other two take up the audio signal and time code. All external system components are built into two 19-inch flight-case racks (Figure 6).

Figure 4. Thick-film hybrid circuit.

Prior to implantation, the prostheses are calibrated for force and temperature measurements in a water basin. Defined forces up to 5000 N are applied. The temperature calibration range is 35 °C to 44 °C. In force accuracy tests, the errors were less than 1% of the maximum calibration force (5000 (spun) N). During the temperature accuracy tests, the errors were below 0.1 °C (Graichen et al. 1999).

Five instrumented hip endoprostheses have been implanted in four patients since 1997. Hip joint forces have been measured during different activities. Additionally, the temperatures along the implant have been determined during walking and bicycle riding. Detailed results will be published soon.

Figure 5. Components of the telemetry system.

Figure 6. Patient and telemetry system.

The instrumented hip joint prosthesis allows long-term in vivo measurements of the hip joint forces and temperatures. All sensors and two 8-channel telemetry units are placed inside the implant and hermetically sealed against body fluids. Each telemetry unit contains an integrated 8-channel telemetry chip and a radio frequency transmitter. Active and passive components are mounted on both sides of a 7 mm x 14 mm wide ceramic substrate using thick-film hybrid technology. The signals are pulse interval modulated (PIM) and transmitted at frequency ranges of 120 MHz and 140 MHz by two antennas. The secondary induction coil inside the shaft of the implant supplies both telemetry units with energy. The magnetic field of the external primary power coil is controlled by a personal computer. Force, temperature, and power supply data are received and demodulated by an external twin-receiver. Force and temperature are monitored in real time and all data are stored on videotape together with the images of the patient’s activity.

This project was supported by the German Research Society (Be 804/11) and by the company LINK (Germany).

Sleep-related breathing disorders are recognized to be very prevalent in the adult male population. The therapy of choice is non-invasive ventilation using a CPAP ventilator at home every night. Alternative treatment is required because acceptance is not higher than 80%. Recently it was demonstrated that electrical stimulation of the upper airway muscles reduces pharyngeal collapsibility, which can reduce obstructive sleep apnea events. A pacemaker has been developed to investigate this effect in humans. The first trials with eight patients worldwide confirmed the positive effects but also revealed a number of technical and clinical problems using this treatment. The new approach appears to be very promising but has not reached the mature stage of routine therapy.

Sleep-related breathing disorders are now recognized as a common health problem. The prevalence in the adult population is reported to be at 4% (Young et al. 1993). The most common sleep related breathing disorder is obstructive sleep apnea. This disorder is characterized by repetitive cessations of oronasal airflow during the night. In patients with obstructive sleep apnea, more than 600 apnea events can be observed during one night. Apneas usually last for 40 seconds to 60 seconds, but apneas as long as 120 seconds also have been observed. Each event is accompanied by drops in oxygen saturation and a cyclical variation in heart rate. During each obstructive apnea, respiratory activity continues while the upper airways collapse. When oxygen drops critically, a cortical arousal activates pharyngeal muscles and reopens the upper airways. When upper airways remain occluded, respiratory activity of the diaphragm and thoracic muscles continues. The increasing effort against the occluded pharynx induces negative intrathoracic pressure swings. Pressure swings, sympathetic drive, and oxygen desaturation also affect the cardiovascular system and may result in cardiovascular disorders.

The treatment of choice today is continuous positive airway pressure ventilation (CPAP). This home ventilation therapy does not lower upper airway collapsibility causally, and therefore has to be used every night to stabilize the upper airways effectively. This principle of a pneumatic splint is a practical therapy and maintains upper airway patency during night. As this device has to be used every night and because of medical and psychological reasons, it is not tolerated by every patient and compliance differs worldwide between 50% and 80%. Therefore, alternative treatments are required that causally lower upper airway collapsibility in order to overcome the limitations of nasal CPAP. Recent investigations demonstrated that electrical stimulation of the upper airway muscles might decrease upper airway collapsibility, improve maximum inspiratory airflow, and possibly decrease sleep-related breathing disorders (Schwartz et al. 1993, Smith et al. 1996). The genioglossus muscle is innervated by the distal branch of the hypoglossal nerve. When this muscle contracts, the tongue moves forward and opens the anterior portion of the pharynx (Schwartz et al. 1998). Therefore an implantable pacemaker for hypoglossal nerve stimulation has been developed in order to take advantage of this phenomenon and investigate its effects on upper airway mechanics during sleep.

Based on experience from phrenic nerve stimulation, we conducted tests involving stimulation of the hypoglossal nerve. A programmable pacemaker (Medtronic Inc.) was implanted in patients with obstructive sleep apnea in a number of selected sleep centers in the world. The general principle of the pacemaker is that it has a lead with a cuff electrode connected to the hypoglossal nerve unilaterally. Additional sensors detect each breath during sleep. Stimulation of the hypoglossal nerve occurs whenever an inspiration is recognized and is maintained during the period of inspiration. The stimulation of the nerve moves the tongue forward (contraction of the genioglossus muscle) and decreases upper airway collapsibility by stiffening the anterior pharyngeal wall. The triggered stiffening and opening of the upper airways prevents an inspiratory collapse and thus may reduce the number of obstructive apneas and hypopneas during sleep.

Sleep and respiration was recorded in our sleep laboratory using polysomnography with EEG, EOG and submental EMG to detect sleep stages. ECG is always monitored as a vital sign and to evaluate heart rate variability. Airflow is measured with a pneumotachograph (Hans Rudolph Inc.) connected to a tightly fitted nasal mask. The air pressure within the mask was measured with a differential pressure transducer referenced to atmospheric pressure. Inductive plethysmography (Respitrace) was used to monitor thoracic and abdominal respiratory movements. Oxygen saturation monitors (Ohmeda, Inc.) and snoring microphone are also standard in the sleep laboratory (Penzel et al. 1993). In these investigations we used in addition esophageal pressure catheters with piezo tips (Gaeltec, Inc.) to quantify changes of intrathoracic pressure during obstructive apnea and hypopnea. Infrared video monitoring and recording allowed continuous supervision of stimulation effects during the nights in the sleep laboratory.

For baseline estimation of the degree of upper airway collapsibility, a critical closing pressure (Pcrit) of the upper airways was determined. This can be done using a standardized procedure when the patient is under nasal CPAP therapy with a pressure high enough to eliminate all obstructive apneas and hypopneas. Applied pressure at the mask and effective airflow have to be monitored carefully. For three subsequent breaths, the pressure is lowered and the resulting airflow is determined. If the upper airway starts to collapse, a limitation of airflow can be recognized according to the physical model of the Starling resistor which is applied to collapsible tubes (Schwartz et al. 1988, Gleadhill et al. 1991). In healthy subjects the Pcrit has values lower than -10 cm H₂O. Typically, snorers have values between -8 and -2 cm H₂O and sleep apnea patients have positive values above 0 cm H₂O. The Pcrit seems to be a characteristic measure which is different in different body positions (supine and lateral positions) and which does not exhibit large sleep-stage-dependent changes.

Patients were selected after nasal CPAP proved to have a beneficial effect but could not be used continuously due to reasons previously mentioned. Patients were actively seeking a possible alternative to CPAP treatment. They were instructed about the scientific character of this study where a therapeutic success cannot be predicted. Patients were selected only if they fully understood the procedure and signed a consent form. The study protocol was approved by the local ethical committees of all centers involved and by the United States Food and Drug Administration. Pacemaker implantation was performed in the department of Cranio-Maxillo-Facial-Surgery with the assistance of an ENT surgeon. Pacemaker settings and clinical investigations after implantation and during follow-up investigations were performed and supervised by sleep physicians.

The first pacemaker developed some years ago was a passive device. It consisted of a pacemaker module with a stimulation cuff implanted at the hypoglossus nerve. The cuff has three contact compartments and is placed around the nerve (Figure 1).

Figure 1. Hypoglossal nerve stimulation cuff placement.

The passive pacemaker module had to be driven inductively through the skin. External sensors for the detection of inspiration were respiratory effort belts based upon respiratory inductive plethysmography. Connected with these inductive belts was an external battery powered device which was programmed to detect every single breath. The device then induced power to the pacemaker, thus causing a hypoglossus nerve stimulation. This device was used to improve the algorithms for the continuous detection of respiration and the proper timing of the stimulus. We also observed stimulus threshold behavior over long time using continuous nerve stimulation every night.

The pacemaker presently used is an active device and has an integrated battery and signal processing unit (Figure 2). The most important feature of this pacemaker is an additional implanted sensor to detect respiration. For this purpose a pressure transducer is implanted in the sternum and continuously records intrathoracic pressure changes as a surrogate for respiration. Thus a "closed loop" stimulation is implemented. To control proper function of the system it is possible to monitor the respiratory signal externally by using inductive data transmission from the pacemaker. This allows the transmission of short signal segments together with markings where the algorithm detected inspiration and expiration.

Figure 2. The system as being implanted in patients with respiratory sensor, pacemaker module, and hypoglossus nerve stimulation.

The principle of nerve stimulation remained the same. The onset of inspiration is detected because the associated step slope is the most distinguishable signal characteristic of the pressure signal. Consecutive calculations of respiratory period and other algorithm timing are based on this detection. The algorithm then performs an automatic gain adjustment of the respiratory signal. A 5% adjustment of gain is possible for each respiratory period; thereby, consistent threshold values for stimulation can be chosen. The hypoglossal nerve is stimulated only during the early part of inspiration. The stimulation is done with a 30 Hz pulse train of stimuli. The maximum duration of the stimulation period can be programmed to be 2, 3, 4 or 5 seconds. A minimum refractory period is maintained after each pulse train.

A total of eight patients were selected for this study with unilateral hypoglossal nerve stimulation. Data are pooled and evaluated using uniform criteria. Here we report about two patients who underwent this procedure in Marburg. In one patient we implanted the passive pacemaker and in another patient we implanted the stimulator with implanted sensing.

Four weeks after implantation a test with selective stimulation of the distal branch of the hpyoglossal nerve resulted in protrusion of the tongue contralaterally, indicating that respiratory sensing, impulse generating, and electrical conduction by the electrode was successful. The tongue protrusion did not disturb the sleep of the patient and did not cause cortical arousals. The stimulation during the night did not affect motor function of the tongue during daytime. Unexpected growth of the genioglossus muscle did not occur.

In the first night of treatment adjustment, the beneficial effect was not the same in all body positions. We recorded again the number of apneas and hypopneas under continuous stimulation and we derived the Pcrit value to determine the improvement of upper airway collapsibility using this new approach. Telemetric data from the pressure sensor showed a smooth waveform pattern representing physiological respiratory pressure swings as expected. Unfortunately, from time to time, the signal showed disturbances. We assume that these disturbances were caused by body movements, by intrathoracic blood vessels, or by cardiac influences.

During follow-up investigations the amplitude of the stimulus had to be increased because the motor threshold for tongue protrusion did increase from 2.8 volts to 3.2 volts. It is possible that connective tissue growth between the nerve and the electrode is responsible for this lower sensitivity.

Compared to the baseline recordings in all follow-up investigations, the effect of electrical stimulation on upper airway patency was not the same in all body positions. In the lateral position maximum inspiratory flow improved by 100 ml/sec to 150 ml/sec, whereas in the supine position airflow increased by less than 50 ml/sec. As a consequence, in the supine position sleep apnea continued to occur at the same rate, whereas lateral position apneas and hypopneas were reduced. Consequently, the effect on the number of respiratory events was body position dependent. In additional ambulatory recordings we could verify that this particular patient spent 70-80% of the time in bed in the lateral position. Thereby, electrical stimulation caused a significant reduction of sleep apnea in this patient. The reduction was as large as 30 apneas per hour of sleep starting with an initial number of 30-50 apneas per hour of sleep. In addition, the activated pacemaker converted apneas with complete cessation of airflow into hypopneas and converted hypopneas into heavy snoring. Although a significant number of respiratory events persisted, clinical symptoms of the patient improved. Figure 3 shows a typical polysomnogram from a patient receiving hypoglossal nerve stimulation during inspiration.

To quantify the effects on upper airway function under continuous stimulation, we derived the Pcrit value to characterize the improvement of upper airway collapsibility. The mean Pcrit value in supine position was lowered from +2.0 cm H₂O to +0.7 cm H₂O thereby indicating no sufficient therapeutic effect in this body position. In the lateral position the mean Pcrit was lowered from –0.8 cm H₂O to –2.5 cm H₂O.

Figure 3. The polysomnographic recording example shows EEG (C3-A2) and EOG on top. Nasal airflow, the derived transsternal pressure signal, thoracic and abdominal movements, EMG submentalis and the stimulation markings follow. On the left side of the recording, an apnea is still present. In the middle of the figure, the patient moves into lateral body position and the continuous airflow is recorded on the right side of the figure. This recording example demonstrates the effect of body position on continuous hypoglossal nerve stimulation.

The new approach of electrical stimulation of the upper airway muscles shows promising results that indicate development of an alternative treatment for patients who suffer from moderate sleep apnea and do not comply with nasal CPAP treatment. The results indicate that this new approach affects upper airway function by decreasing its collapsibility. This effect can be quantitatively measured as a shift in critical pharyngeal closing pressure. Assuming that electrical stimulation of the genioglossus muscle might lower Pcrit about 5 cm H₂O, it is therefore obvious that hypoglossal nerve stimulation cannot diminish apneas in patients with severe sleep apnea and high Pcrit values (> 5.0 cm H₂O). In those patients obstructive apneas will remain unaffected. In these cases, apnea would persist and would still cause oxygen desaturation and cardiovascular sequela.

The multi-center study also revealed a number of technical problems. Our pressure sensor failed to work after twelve months due to breakage of the nerve cuff. Obviously physical stress on the cuff is higher than expected. The algorithms for detecting the inspiration also are subject for improvement. Movement artifacts and probably pressure changes induced by the heart were the most important problems observed. To minimize these influences it might be advantageous to change the location of the respiratory sensor. This would also help to reduce the risk of the surgical procedure. These different problems with the present system initiated the development of a new, improved version that will be tested in clinical trials beginning at the end of 1999.

A number of physiological questions also remain open. It still is not clear how much effective stimulation is dependent on sleep stage. This question is closely linked to upper airway anatomy changes under different sleep stages. The importance of body position during sleep on effective stimulation became obvious in our patient. These effects need systematic evaluation. And finally, it is not clear which patients might benefit most from this new treatment approach. Certainly rules for Pcrit values and for number and type of apneas have to be established before a general recommendation can be made.

We want to thank the Upper Airway division at Medtronic Inc, Minneapolis (USA) and Maastricht (The Netherlands) for their partial support of this study. Results of the international multi-center study are being used today to improve the stimulator for a second generation.

A PIC16LC84-based implantable stimulator was used to examine the roles of aggregate impulse activity and force in the adaptive response of mammalian skeletal muscle to increased use. Constant-frequency stimulation and an optimized stimulus pattern producing three times the force had similar physiological and biochemical effects, but differed in preservation of muscle mass and activation of contraction. The results suggest that the initiating events belong to a stage preceding force development.

Chronic electrical stimulation of a mammalian fast skeletal muscle over a period of weeks produces profound changes in structure, contractile speed, and metabolism. These changes are adaptive; they enable the muscle to support a sustained increase in activity without fatigue by decreasing the energy needed for contraction while increasing the capacity for providing that energy via aerobic metabolic pathways . The changes involve both quantitative and qualitative changes in the protein composition of the muscle, and many of these are the result of regulatory events taking place at the level of the gene. However, the cellular signalling pathways that bring about these changes in gene expression are unknown. In broad terms the process could be initiated either by electrical events and the associated ionic movements, or by force generation and the associated energy expenditure. This study was designed to distinguish between these possibilities.

We compared the effects of two different patterns, delivered without interruption over a period of 12 weeks. One consisted of constant-frequency stimulation at 2.5 Hz. Previous work from this laboratory has shown that this pattern induces transformation into an essentially "Type 2A" muscle, which combines high fatigue resistance with fast contractile characteristics. The other pattern delivered the same aggregate number of impulses but with the maximum possible difference in the tension generated. This was achieved by choosing the interpulse intervals (IPIs) within each burst so as to maximize the mechanical output per impulse . This optimized stimulation pattern (OSP) utilises the "doublet effect," a more-than-linear summation of tension that occurs at short IPIs (see and references cited there). The doublet effect on tension-time integral is diluted as increasing numbers of impulses are added to a burst, and the greatest departure from constant-frequency stimulation is seen when the impulses are delivered in bursts of three . The chosen OSP was, therefore, a triplet burst pattern, for which the isometric force-time integral was more than three times the equivalent of three individual twitch contractions.

The technical challenge of generating this complex pattern with an implantable stimulator was met by using a device based on a PIC16LC84 microcontroller . This article focuses not so much on the device as on the type of scientific work that it makes possible.

Constant-frequency stimulators were constructed according to a previous design , with a resistance-capacitance combination that set the frequency to 2.5 Hz. The pulses were of 0.2 ms duration and 3.2 V amplitude. OSPs were generated by a new device based on a commercial microcontroller (Microchip PIC16LC84) . The implant was 20 mm in diameter and drew less than 40 µA operating current from a single integral 3V lithium coin cell. Twelve different stimulation patterns were stored in the EEPROM data area of the PIC, including a sleep or off mode, and the required pattern was selected after the device had been implanted by directing a sequence of high intensity light flashes through the skin . The flashes were detected by a phototransistor in the implant and counted by the processor; the selected stimulation pattern was then initiated. The OSP consisted of a burst of 3 impulses, with IPIs of 6 ms and 12 ms, delivered every 1.2 s. This delivered the same aggregate number of impulses as the constant-frequency stimulator, nominally 150/min.

The experiment was conducted on 12 adult New Zealand rabbits. Commencing one week after the implantation operation, the tibialis anterior (TA) muscle of the left hind limb was stimulated continuously and supramaximally via its motor nerve for 24 h/day, either with 2.5 Hz (n=6) or with the OSP (n=6). The right TA served as an unstimulated control muscle. After 12 weeks, physiological measurements were made in a terminal experiment. Details of operative and measurement procedures may be found in a previous publication . In addition, the doublet effect was assessed by recording the contractions elicited by two impulses with IPIs between 1 ms and 20 ms. Finally, each muscle was subjected to a 30-min fatigue test.

The fiber type composition of samples from the TA muscles was determined in cryostat sections stained histochemically for the demonstration of myofibrillar ATPase. Other samples were used to determine myosin light-chain and heavy-chain composition by SDS-polyacrylamide gel electrophoresis.

Stimulated muscles were compared to their contralateral counterparts by a paired Student’s t-test. To assess the difference between the responses to the two patterns of stimulation, the arithmetic differences between the stimulated and contralateral muscles in each group were compared by the Mann-Whitney non-parametric test.

Stimulation resulted in a reduction in contractile speed that was similar in extent for the two patterns, maximum shortening velocity and power (Figure 1). Twitch contraction time was prolonged significantly for both patterns of stimulation, but the effect of constant 2.5 Hz significantly exceeded that of the OSP; a similar pattern was observed for the twitch:tetanus ratio. The doublet effect was still substantial after stimulation with the OSP, but barely detectable after constant 2.5 Hz. Muscle mass decreased more under the constant-frequency regime than for the OSP, and the difference was significant (Figure 1).

Figure 1. Changes in selected physiological properties of rabbit TA muscles induced by stimulation for 12 weeks at 2.5 Hz or with an equivalent optimized stimulation pattern (OSP). Changes are expressed as the percentage difference between the stimulated and the contralateral control muscles. With the exception of the twitch:tetanus ratio in the OSP group, the variables shown all departed significantly from control values. The asterisk (*) symbol denotes a significant difference (P<0.05) between the two types of stimulation.

Both of the chronically stimulated groups of muscles were much more resistant to fatigue than the contralateral control muscles (results not shown). The two stimulated groups did not differ significantly in this respect.

Chronic stimulation significantly increased the proportion of Type 2A fibers at the expense of Type 2D fibers. As a result, Type 2A fibers became the predominant fiber type in both stimulated groups, and there was no significant difference between them (Figure 2). The small increase in the proportion of slow, Type 1 fibers did not achieve significance. These changes were also reflected in gel electrophoretograms of myosin isoform composition (results not shown).

Figure 2. Fiber type composition of control muscles and muscles stimulated for 12 weeks with the OSP or with 2.5 Hz.

The original implantable muscle stimulator was developed to address the question: what is the role of activity in the differentiation of fast and slow skeletal muscle? That simple device made it possible to alter the impulse traffic reaching a muscle over a period of many weeks and led to the discovery of the adaptive capacity of skeletal muscle . The latest generation of devices now allows us to generate more sophisticated patterns and so to address some of the many scientific questions that are still outstanding in this field. Such devices can be designed around Application Specific Integrated Circuits, including gate arrays , but the microcontroller approach is more flexible and less costly, and features can be exploited that make the devices fault-tolerant .

For many years it has been our working hypothesis that mammalian skeletal muscle responds adaptively to the aggregate amount of impulse activity it receives from the motor nerve . According to this hypothesis, fibers subjected to sustained high levels of use tend to develop properties at the slow, fatigue-resistant end of the spectrum. Fibers that are less active retain, or revert to, a native fast state. The ability of the fibers to adapt continuously to their pattern of use would allow a muscle to respond to changing functional demands throughout adult life. This hypothesis explains not only the fiber type transitions seen in increased use, brought about, for example, by chronic stimulation, hypergravity, and endurance exercise, but also the reverse transitions seen in decreased use and disuse, as in hindlimb suspension, zero gravity, joint fixation, tenotomy, and paralysis due to spinal cord injury. It differs from formulations of the activity hypothesis that place the emphasis on the frequency of impulse activity rather than on its aggregate amount.

If frequency, as opposed to aggregate activity, had been an important differentiating influence, we would have expected stimulation with the OSP to produce smaller changes in shortening velocity, myosin isoform composition, and fiber type histochemistry. It did not. Aggregate impulse activity was the only feature common to the two patterns, and must therefore have been the major initiator of changes in these properties. Force can be excluded as an initiator of changes in contractile speed, because the OSP produced at least three times the force-time integral of the constant-frequency pattern. Changes in oxidative capacity (reflected in resistance to fatigue) did not differ between the two groups of muscles, and were probably related to the energy cost of the activity rather than its precise pattern .

So in what ways did the two patterns differ in their effects? First, the OSP was more effective in preserving muscle bulk, suggestive of a higher level of protein synthesis in these muscles. The greater forces elicited by the OSP were probably responsible for this effect, as force is known to be a powerful general stimulus for protein synthesis. Second, changes in time to peak twitch contraction and twitch:tetanus ratio were far smaller after the OSP than after constant-frequency stimulation. This points to differences in activation between the two groups of stimulated muscles, a conclusion that is strengthened by the changes in the doublet effect. Calcium transport must therefore be regulated separately from properties such as shortening velocity.

In conclusion, this study has shown that although muscle force is a potent general regulator of protein synthesis, the signals that regulate the expression of the individual genes involved in adaptive change are associated with an earlier stage of muscle activation, such as membrane depolarisation or the release and reaccumulation of calcium ions.

Department of Engineering, University of Tras-os-Montes and Alto Douro,
Apt. 202, 5001 Vila Real, Portugal

Department of Biological and Environmental Engineering,
University of Tras-os-Montes and Alto Douro, Apt. 202, 5001 Vila Real, Portugal

Accurate heart rate monitoring was achieved through heart sound sensing using a promising subcutaneous implantable device. This paper presents the capsule mechanical design, piezo sensor theoretical and practical responses, analog heart sound signal processing, and results from mild subcutaneous implants for heart rate variability studies carried out in Sheep. Simultaneous electrocardiogram and acoustic heart rate waveforms were recorded in order to validate the heart sound sensor and signal processing methodologies.

Animal physiological responses to stressful conditions can be followed through changes in heart rate. Heart rate variation is one of the meaningful indicators of animal welfare. Sheep show clear heart rate responses to the approach of dogs, to introduction to a new flock, to being visually isolated from their companions, to being loaded onto a truck and during transportation. Heart rate variability monitoring offers a convenient way of assessing animal housing welfare, and short term welfare problems during handling and transportation, which usually result in significant economic losses (Fraser and Broom 1998). On the other hand, tachycardia accompanies fever, inflammation, pain, hypocalcemia or metabolic disturbances that result in hypovolemia, and bradycardia is seen with heart conduction disorders within the heart muscle and in uraemia and in hypokalemia. Frequent telemetric heart rate monitoring can be widely applied in animal production at the farm level, being an affordable and reliable means of assuring humane animal management practices.

Biomedical applications of piezoelectric materials include ultrasonic wave detection, using sensor arrays, microphones in the audible range for arterial Korotkoff sound in blood pressure detection, and even piezofilm shoe insoles for the evaluation of the kinetic variable of human locomotion (Nevill et al. 1995, Shung and Zipparo 1996). Piezoelectric sensors also detect on the chest surface the acoustic vibration produced by the heart’s mechanical activity. Considering this, a small dimension sensing tube and an analog signal processing device were developed for heart rate monitoring (Torres-Pereira et al. 1997a). A most convenient way to couple electronic circuits to elastic waves relies on the use of the piezoelectric effect. This results in the production of electric polarisation charge when a piezoelectric crystal is stressed. If the applied stress is maintained for a while, this charge will be neutralized by internal leakage. Thus, a piezolelectric sensor responds to a sudden stress rather than to a steady level of applied force (Fraden 1996).

Piezoelectric properties exist in some crystals, such as quartz (SiO₂), in artificially polarized ceramics (such as BaTiO₃) and in some polymers (such as PZT). Ceramic materials have several advantages over single crystals, such as higher sensitivity (up to several hundred times higher) and ease of fabrication in several shapes and sizes. A major advantage of PVDF piezofilm over piezoceramics is its low acoustic impedance, closer to that of animal tissue. A good impedance match allows better transduction of acoustic signals. To pick up electric charge, conductive electrodes must be applied to the crystal at the opposite sides of the cut. As such, a piezoelectric sensor is a capacitor whose dielectric material is piezoelectric (Fraden 1996).

PVDF film is typically thin, flexible and sensitive, thus well suited to strain sensing applications requiring wide bandwidth and high sensitivity. Amplification of PZT actions can be achieved by various shapes, like tube, thin disk, and plates, in unimorph or bimorph arrangements. An unimorph structure is achieved by bonding a thin piece of piezoceramics to an inactive substrate, such as in hydrophones and in vibration actuators. A bimorph structure is obtained by bonding two plates (films) of piezoceramics together, so that differential changes in the length of both plates can originate significant, detectable, movement, such as in our acoustic vibration sensor.

The sensor used is 15 mm long, 1.5 mm wide and 0.6 mm thick. The piezoelectric constant and its capacitance are 12.1x10^-3 VmN^-1 and about 750 pF, respectively. Its electromechanical coupling factor is 60 and its compliance is 6.6x10^-4m/N (RS 1997). The sensor is a bimorph piezoelectric sensor glued with non-conducting epoxy to the internal surface of a tube. This piezo ceramic bimorph element is a self generating electromechanical sensor. When the sensor is subjected to vibration, the minute movement causes one layer to be under tension while the other is under compression. Since the two layers are polarized in opposite directions, the opposite stresses originated in each layer give rise to a double output signal. The capsule tube is of polyallomer, a translucid, non-biodegradable linear copolymer material with repeated sequences of ethylene and propylene, similar to that of permanent surgical threads. Originally it was a centrifuge tube, with good mechanical strength and capable of withstanding temperature cycles.

Invasive and non-invasive tests led to the design of analog signal processing electronics based on a TLC27L9 ultra-low power, four operational amplifier chip. Its functional blocks are a charge amplifier, a second order active low-pass filter implementing an envelope detector, an amplifier with automatic gain control (AGC), a threshold circuit based on a peak detector and, finally, a comparator (Figure 1).

Figure 1. Block diagram of the analog signal processing module.

The charge amplifier is used as a signal conditioning circuit for the self-generating piezo sensor and it plays a key role in the piezo sensor performance. A charge amplifier is a circuit whose input impedance is a capacitance that provides a very high value of impedance at low frequencies. Contrary to what its name may suggest, a charge amplifier does not amplify the electric charge present at its input. In fact, its function is to deliver a voltage proportional to such charge and to provide it at a low output impedance. Hence it is a charge-to-voltage converter. The frequency range and signal amplitude requirements over the desired dynamic range must be taken into account in its design (Pallás-Areny and Webster 1991).

A device for long term, short range, telemetry was developed. The electronics are housed in the polyallomer tube (dimensions 75 mm x 17 mm) that encapsulates: a) the heart sound rate detection circuit (25 mm x 15 mm), b) a miniature FSK transmitter (22 mm x 15 mm), c) a 3.6 V battery (28 mm x 14 mm), and d) an antenna. The antenna coil was wound around the battery to reduce the capsule size. The piezoelectric sensor was glued with epoxy to the internal surface of the capsule (Figure 2).

Figure 2. Implantable Capsule.

We tested our PCG device in a sheep in close-to-natural conditions, i.e. monitoring/testing took place in the usual restraint area. The animal was a 50 kg young male sheep. There was no need for food/water restriction since the animal was not placed under deep sedation or general anesthesia. The electronics equipment was close to the restraint area. The animal was held with a collar. Cardiac auscultation was performed first with a stethoscope, then with a sensor placed inside the capsule (polyallomer centrifuge tube), to detect: a) where heart sounds were stronger in the animal (mitral area - S1), b) where externally the S1 signal was better detected by the sensor, and c) where it was adequate to place the capsule to prevent eventual forearm-trunk friction.

A small surgical area was shaved and washed. The incision site was cleaned with iodopovidone. Local anesthesia was performed with 2% procaine (infiltration of a small circular area). External PCG tests were performed again with our device, to assure the right location of the capsule implant. A small incision the diameter of the capsule was made in the skin of the surgical area. Subcutaneous tissue was displaced with light circular movements, helping the capsule to be easily implanted subcutaneously. Since we were testing a prototype, the incision was not closed. At the end of trials, the capsule was removed and the incision sutured.

Several PCG/heart rate recordings were performed during a two-hour period, the animal being relaxed in the presence of our team and the stables workers. It is important to note that such good quality signals were obtained without the administration of any tranquilizer to the sheep. In the same animal, classic external ECG tests were conducted at the same time in order to validate the PCG signals under close to field conditions.

After these trials, the animal was released from the portable restriction area, the incision being closed with two isolated stitches. After intramuscular administration of penicillin, the animal was returned to its flock. It has been followed and it is so far in an excellent condition. Since the capsule used proved to be quite good for in vivo PCG data transmission, a small group of sheep will be used to test its long term biocompatibility for permanent implants.

A stethoscope was used to locate the loudest heart sound point on the animal’s thorax. The strongest S1 signals were obtained lateral to the apex of the sternum on the left side of the animal. Sound signals at the precordial area were considerably lower. These facts were confirmed by the capsule piezoelectric sensor. During digestion, heart sounds recorded at the traditional precordial auscultation area have lower noise. The envelope detection of S1 and the signal processing module were the waveforms recorded when the capsule was subcutaneously implanted in the precordial area of the animal (Figure 3).

Figure 3. S1 envelope detection and digital output from an implanted capsule.

In order to validate the heart sound sensor and signal processing methodologies, the acoustic heart rate waveforms were compared with heart rate monitoring by ECG (Figure 4).

Figure 4. Simultaneous PCG and ECG recordings.

Simultaneous recordings were carried out over fifteen minutes. Comparison of these two recordings showed a slight delay of the S1 peak wave over the R peak of the QRS complex. In fact, the depolarisation process of the heart detected by the ECG signal triggers a wave of contraction spreading through the myocardium. Atrial systole starts after the P wave, and ventricular systole starts near the end of the R wave and ends just after the T wave. The first heart sound (S1) occurs at the onset of ventricular contraction during the closure of the mitral valve, indicating the beginning of ventricular systole (Torres-Pereira et al. 1997b). The delay caused by low pass filtering can be neglected. The QRS complex and the acoustic S1 envelope waveforms have equal frequencies.

These results validate the use of the heart sound rate sensor approach as a means for telemetric heart rate variability studies.

Accurate heart rate monitoring of heart sounds (PCG) was achieved using a subcutaneous implantable device. This strategy is successful because it does not require external wiring, the system being easier and safer to implant when compared with ECG. On the other hand, the costs of piezocontact microphones are one order of magnitude lower than those of microphones based on accelerometers. Air-coupled microphones are sensitive to undesirable environmental noise and would be useless inside an enclosed capsule, since no air pressure gradients are present.

Further advantage of piezocontact sensors, which is of paramount importance in implantable devices, derives from the production of sound waves with no power consumption. The sensor is part of a data-acquisition system, which is, in turn, part of an electronic measurement system to acquire information, without interference from any undesired signal source. These conditions are, of course, only met up to a certain extent in a practical sensor. Signal conditioning circuits were implemented to reduce these effects by offset cancellation, output impedance transformation, amplification, and filtering.

The present analog signal processing electronics development stage, as well as its in vivo tests using an implantable capsule, clearly point to the design feasibility of a small capsule based on the joint encapsulation of a PCG heart rate module together with a microcontroller and a telemetry link (Torres-Pereira et al. 1997c, Geeraerts et al. 1997). Reducing data processing overhead on the external computer by analog signal processing at the sensor site, thus making it self-adjustable to physiological signal variation, and using digital transmission are feasible solutions.

Signal stability indicates that our PCG device can be used in continuous monitoring for studies on heart rate variability both in humans and in animals. In our hands, heart rate monitoring based on PCG has the advantage over heart rate monitoring based on ECG of only requiring a single sensor probe and of conveying a more stabilized signal under close-to-field conditions. Telemetric heart rate monitoring based on S1 detection can be widely applied in animal production at farms, being a means to improve both health surveillance and animal welfare. We predict that adding the information obtained from our acoustic heart rate monitoring system to that coming from temperature and activity sensing devices will provide more adequate understanding of farm animal behavior physiology, e.g., reproductive cycles, and therefore a safer management of animal production.

The authors are grateful to the European Union for funding this work, which is part of Research Project EEC-CT94-2304 (CAPT: Coupling Active and Passive Telemetric Data Collection for Monitoring, Control and Management of Animal Production at Farm and Sectorial Level), Research and Development Program for Agriculture and Fisheries of the European Union.

SICAN F&E GmbH, Advanced Systems, Hannover, Germany, and
Technical University of Berlin, Institute of Electronics, Berlin, Germany

The possibility of pressure and temperature measurement in all areas of the intensive care unit (ICU) with mobility of the patient, would be a great benefit (Gaab and Heissler1984). Our work shows how a fully implantable stand-by device for measuring intracranial pressure and temperature in the ICU, and under everyday conditions, can be implemented. It consists of a sensor element (Kersjes and Mokwa 1996) combined with a transcutaneous telemetric interface (Jurisch et al. 1993).

Investigations of intracorporal pressure and temperature measurement have become very popular (Talamoti et al. 1988), largely due to the increase in available microsystem components (Bryzek 1992). There are many advantages of the implantation of a telemetric unit (Figure 1) instead of measuring by a catheter (Chapman et al. 1990). One of particular importance is the simplification of patient care, so some systems are working absolutely batteryless. If the extracorporal coil is not connected, the system will only be a piece of silicon, packaged in silicone. It will awake when it is necessary (Jurisch 1991).

There are several methods for measuring the intracranial pressure (ICP). One of particular interest is the direct measurement in the liquid-filled cavities (intraventricular) of the brain (Figure 2). This placement allows very precise measurement, because of its direct liquid coupling.

Figure 1. A passive stand-by system for telemetric transmission of physiologic information.

Figure 2. The sensor locations for measuring the intracranial pressure (ICP).

The disadvantages of brain cavity measurement are: 1) the risk of bleeding in the brain due to damage to an artery or other vessel, and 2) the risk of damaging the brain tissue or causing meningitis. It is also possible that the brain cavities are collapsed because of a chronic brain disease. In this case, an intraventricular ICP measurement is impossible (Czosnyka et al. 1996).

For routine and chronic measurements, it is only possible to measure epidural pressure, but this has the disadvantage of an indirect method, because of the transmission via the dura mater (B.-Flick et al. 1996a, 1996b). This method has the advantage, however, that it is impossible for the sensor to become imbedded in the brain tissue.

The signal from the human body is sampled via a portable data recorder, consisting of a PIC16C61-controller to convert the demodulated PWM signals into a serial bit stream (Figure 4).

Figure 3. The telemetric system.

Figure 4. A master-slave solution for data-management.

The data sampling controller is linked by a RS232 connection to a second controller (Intel 80C517) that collects the input data (intracorporal and atmospheric temperature and pressure), shows them on an LCD-display, and stores them on a PCMCIA-SRAM card (Figure 5). These cards are compatible with laptop standards and can be easily changed, if they run out of memory. The sampled data can be transferred via postal service or via modem and telephone to a computer at the hospital. In the future, this kind of transmission, called home monitoring, will significantly reduce costs in the medical field.

Another point of interest is automatic event recognition, which is used to change the sampling rates in order to capture special signal components in an emergency situation. Therefore, a signal processing and waveform analysis is necessary, first to control the measured signal in real time on the portable unit, and second, to process the data off line on the stationary unit (laptop) to calculate the compliance and the resorption of the brain.

This new measuring system could also be implemented in other medical fields, where telemetric sensors are needed and biological signals up to the 50 Hz range are measured. Such applications would be, for example, a stand-by device for the measuring of glucose for diabetes mellitus or monitoring/observing higher risk cardiovascular operations, diseases and events.

Figure 5. Demonstration of the telemetry system.

We thank the SICAN F&E GmbH and the NIGAN e.V. for their material and financial support, which helped to build this system. In addition, Véronique Chastrusse, Agnès Biret, Christian Beck, Bertrand Cerou, Marc Dörpmund, Guido Eckhard, Olivier Berque, Franck De Clerck, Thomas Metzler, and Hans-Jörg Zietz helped to realize our ideas with their diploma theses. Also, many thanks to Hans E. Heissler of the Medical University of Hanover for his engaging discussions, and to Britta Probst and Carsten Mehnert, who helped to test and specify the developed epidural measuring system.

CNR Institute of Clinical Physiology, Pisa, Italy

A detailed description of a real application of the basic technology already developed by the Authors for telemonitoring in ward management is presented here. Some of achievements and devices designed for personal ambulatory monitoring have been dedicated for bed side (bio)-signal logging, and part of the software manufactured for tele-control have been used in a sophisticated management program able to control and monitor activities in medical departments (e.g. cardiology department). The adaptation of the multi-function telemonitoring terminal (presented at the 1997 International Symposium on Biotelemetry in Marburg, Germany), resulted in a simplification of the device with respect to the communication section (the telephone transmission facility has been substituted by a galvanic and /or radio method) and in the development of special front ends to detect and collect signals in addition to EKG. Additional variables such as an oxygen saturation measurement system, the interface to on/off alarms such as those generated by the infusion pumps, the digital link to automatic blood pressure measurement devices, etc., are now included. The resulting unit is quite different in terms of complexity due to the design and implementation of the control unit.

The increasing diffusion of computer technology and telecommunications has made the telemedicine approach, the providing of medical information and services at a distance a fundamental tool to harmonize quality and cost in the healthcare system (McCue et al. 1998). While the most common use of telemedicine has been, up to now, in the long distance links between the patient and the physician, the object of this work is a hospital application in which both the technologic and the management aspects have been addressed. Information technology allows the acceleration of the information processing, offering easy and quick procedures to record, archive, analyze and search the desired data. The hospital represents a complex organization that requires the control of different kinds of data for the management of personnel, patients, clinical information, etc. Only effective management is able to guarantee coordinated and integrated activity of the different parts of the system: the more complex the organization, the more effective the information system must be. Many transcriptions may generate therapy errors; insufficient communication between different professionals may generate misunderstandings; and incorrect monitoring of drugs may lead to serious consequences. Finally, from the management point of view, only the availability of correct information about the workloads in the different units of the hospital makes it possible for effective distribution of human resources.

All these goals may be achieved by introducing an information system. This requires considerable changes in the organization. To completely exploit the possibilities of the system, every information transmission, every process and activity has to be rationalized, making the information easily and immediately available for the people who really need it. The problems to be faced in the setting up of the hospital information system are mainly related to the level of acceptance of the involved personnel, who are forced to change their work routines. Another thing to bear in mind is the necessity to assure the effective ability to guarantee the data security (Goldberg et al. 1998). The problem of the personnel has emerged as a crucial one in ward work, where the nurses, already engaged in offering important, often critical care, have to measure themselves with a new tool about which they normally have no specific knowledge.

The nurses have their own goals and so need specific information to define them and set up the procedures necessary to accomplish them. By monitoring these targets and processing the related data, the improvement of information available to the ward environment may eventually result in better care offered to the patients. The set up of the Ward Informative System (WIS) has required a standardization of the phases of the nursing- care process, an identification of the truly significant data, and a codification of the patient needs and nurse interventions. This work has been carried out in the context of a more general and ambitious project of the Italian Health Department for the development, set up, experimentation and evaluation of an integrated system for the optimized management of the resources in the field of cardiovascular diseases (SPERIGEST). The main feature of the general project has been the search for a global approach for the integration of the sanitary, the technologic, the epidemiological and the administrative components, traditionally not linked in the flow of the data.

The WIS has been set up in the Institute of Clinical Physiology of Pisa, in three wards, two of cardiology, intensive and subintensive, and one of pulmonology, with a total of 33 beds (16 intensive). Twenty-eight nurses work in the wards. The most frequently observed pathologies are: myocardial infarction, angina pectoris, arterial and pulmonary hypertension, congestive and hypertrophic cardiomyopathy. The mean patient stay is 7 days.

The WIS is composed of three main components: the surveillance monitor system (Figure. 1), placed at the patient’s bedside, the personal datalogger (Figure 2) and the central control station (Figure 3). The authors have previously reported about the hardware and the software connection between the monitoring line and the central station (Franchi et al. 1998).

The hardware surveillance system at the bed side consists of a proprietary Pentium based PC computer with a touch screen user interface, while the signal logging device consists of a proprietary palm-top datalogger, based on a Hitachi 16-bit RISC microcontroller of the H8/500 family. The developed datalogger is characterized by the following features: presence of an on-board converter, eight general purpose analog channels, 1 Megabyte static random access memory, and high speed serial channel. To obtain a more flexible and modular acquisition system, the analog front-end has been split from the digital core of the datalogger. Special front ends have been developed to continuously detect ECG and oxygen saturation. The system surveys the on/off alarms generated by the various infusion pumps. Commercially available digital sphygmomanometers have been connected to the hardware surveillance system. The designed devices themselves constitute a system that is able, failing the link to the central control station, to directly monitor the physiologic signals acquired at the bedside. The central control station is based on a Pentium PC and is connected to the ward beds and to the Institute general archive continuously monitoring the ward activity and exchanging the desired data with the general archive.

Figure 3. The nurse ward control central station.

Our systemic approach has led to the development of a central control station that is able to work separately from the general archive if the physical link fails. In order to make it possible to integrate the WIS with peripheral archives, installed on different hardware and software platforms, particular attention has been devoted to the phases of the exchange of data through the network, directly with remote databases.

The software of the central control station has been developed in Visual Basic, resulting in an easy to use interface. The local database on the WIS has been developed in the Access environment.

Focusing our attention on the real work of the nurses, we established a standardized set of procedures, and then modelled of the nurse activities that have been implemented in the management part of the central control station. The exact location of the patient in the ward and their gender are immediately visible on the screen, while the information about the patient status and general procedures of the ward are inserted and read on specific forms (Figure 4 and Figure 5).

The patient’s temperature, heart rate, blood pressure, urine output, weight, etc. and the time this information was collected are quickly and easily written on a form to record and store the clinical situation of each patient. The results from a clinical examination and all the necessary information to carry out the actions relevant to the patient care are entered on specific forms designed for that purpose (Figure 6 and Figure 7). The ordering and availability of drugs are monitored in a special form, allowing an on-line control of the available provisions.

The developed system has demonstrated several advantages for the physicians and nurses in patient management and to harmonize clinical needs and resources. The system helps the nurses by providing: a quick and safe control of their work, the delegation to the PC of tedious tasks, the capability of easily managing a large quantity of data, the quick exchange of information with the other members of the hospital organization, and the standardization of procedures.

The physician is supported in the clinical evaluation of patients by the great amount of data automatically inserted "on-line" in this system, by the hardware surveillance devices and by the information on the direct nursing intervention. All the information about the use of beds, the supplied drugs, and the kinds of procedures more frequently performed by the personnel has made the administration able to optimize the available resources to meet the needs of the patients. A fundamental step for the success of the implementation of the WIS, has been the involvement of the personnel themselves in the set up of the system, not the preparation of the un-skilled personnel. This work has been performed with an effective and fruitful co-operation of the nursing and medical staff and generated a product that, in addition to the true management mechanization, provides the basis to associate, catalogue and quantify the patient needs, the interventions and the results. In fact, the collected statistics in three months of operation in a cardiology department produced about 80 registrations per patient per day; this is useful to generate a large and exhaustive database to investigate in the field of the modern nursing technique, focused to increase the patient service and to lower costs, maximizing the efficiency/intervention ratio. The present work, together with a detailed hardware and software component description, leads to results also useful in terms of department operation. These results otherwise will be difficult to collect without a systematic, rational, quantified and global monitoring of the activity.

Istituto Superiore di Sanità, Roma, Italy

CNR Institute of Clinical Physiology, Pisa, Italy

Dept. of Internal Medicine, University of Pisa, Italy

ISIA, Florence, Italy

A number of multi-sensor devices are being proposed to monitor patients at home, subjects at risk during activities of daily life, and athletes during long-lasting activities. Among potential applications, the surveillance of patients at home is particularly attractive for the health services, as it would reduce hospitalization time and costs. A requirement common to all these applications is the reduction of device encumbrance to a minimum, which is important both for the acceptance by the subject and for not affecting measurements significantly. These devices should ideally remain out of view in the case of patients, or on the contrary, they can be visibly worn by body-fitness athletes.

Three main considerations have guided the design of our device: it should have a prosthetic dimension, it has to be non-invasive in a broad sense, and has to be user friendly, i. e. act as a natural extension of the human body and/or of its sensory capacities. The main technical requirements were: miniaturization, limited weight properly distributed over the body surface, and modularity to adapt the device to different applications. Device components are: a user’s interface; a control unit based on a RISC processor with 24 MB of logging memory that guarantees 24 h of continuous operation; sensor interfaces for EKG, blood oxygen saturation, accelerometers, pressure insoles, an ultrasonic distance meter; an FM 433 MHz radio modem and/or a GSM cellular phone.

The architectural constraints were taken into consideration from the early stages of the electronic design. The main components were built into different units. The user’s interface has a 4 x 4 keyboard and a 64 x 128 pixel graphic display. The device can be constituted by maximum six units, in addition to the control unit. All the units are connected through a single bus, manufactured on a flexible kapton support with the shape of a short belt that carries equally spaced plug-in connectors. The assembly, total length about 35 cm, is then inserted in a special, wearable harness. The assembly is fastened over the abdomen, just under the lower ribs, by means of this brassiere-like harness. It has been made in different fabrics and colors to fulfil the aesthetic requirements of the different applications. The cables coming out from the harness connect the sensor units to the sensors placed on different sites of the body surface: thorax, and upper and lower extremities, including feet. This harness arrangement has proven to be effective for easily positioning the device, for distributing weight over the body surface, and for rendering the device unnoticeable under clothing. The principal result is that the device is almost unperceived by the subject. This is particularly important since its main scope is the monitoring of motor activities and the evaluation of motor capacity in the elderly.

The ever growing demand for health in an aging world together with the increasing cost of hospital care has made physicians, engineers, and economists focus their attention on the potential advantages of telemedicine. The approach offered by telemedicine has the main advantage of conjugating resources and demands, preserving a good quality of life for patients while containing the economic impact on the national health systems. The important technological achievements of information and telecommunication sciences have made it possible to extensively exploit the telemedicine approach today. Thanks to telemedicine, post-hospitalized subjects, chronic patients, and the elderly can be, and feel they are, safely assisted even at home, while physicians are offered the possibility of an early and more effective intervention. A better quality of life is obtained for both patients and their relatives, not to mention the reduction of hospitalization time and, consequently, the costs for the health systems. The work described in these pages has been carried out following this approach. A multi-sensor miniaturized device has been developed for monitoring of elderly and chronic patients out of the hospital environment. The device also has been found suitable for monitoring athletes and healthy people engaged in competitions or in normal physical training activities (Figure 1). A requirement common to all these applications is the limitation of device encumbrance to a minimum, which is important both for the acceptance by the subject and for not significantly affecting the measurements.

Figure 1. Sketch of the multifunctional telemonitoring architecture.

From the very beginning, the device was conceived as featuring a high degree of technology integration. It consists of 3 interchangeable modules, in order to assure the necessary flexibility in answering to the needs of the specific application (Figure 2). The following clinical parameters are sensed and acquired: ECG, 2 D acceleration of the trunk, inter-foot distance, ground reaction force and its point of application for each foot, and, when necessary, blood oxygen saturation. The core of the device is a digital controller with the following basic features: a RISC, C-native, 8 analog (10 bits) channel processor, equipped with an onboard RAM (1 MB), an RS232 serial interface (50.000 baud), and a real-time clock. The user’s interface module consists of an LCD graphical display (64 pixels x 128 pixels), a 6-key keyboard, and a buzzer.

Figure. 2. Main components. From left to right: accelerometer interface, ECG and Oxygen Saturation meter front end, Digital Control Unit, inter-foot U.S. meter, Memory Card.

We developed and tested both the hardware and software of the telemetric link; the FM standard 433 MHz radio used for short range applications (100 m) can be replaced by a dedicated GSM cellular phone module when a longer range is required. Special care was devoted to the design of the assembly system for all the sub-units over the body. As sketched in Figure 3, the harness carrying the electronics resembles a brassiere. The cloth belt positioned just under the lower ribs contains a proprietary wiring bus manufactured on a kapton support. Plug-in connectors assure the link with the various modules and on-body remote sensor units.

The bus was designed to minimize the electronic interfaces with the modules, in the form of electronic cards, and to exploit the kapton support to accommodate a relatively high number of wires. The 64-wire bus is subdivided into two parts: a 54 wire digital bus and a 10 wire analog bus that can accommodate up to eight analog channels. The two buses are located in two different lanes separated by a ground strip. The connection of an electronic card is performed by means of two 32-pin (2x16) connectors. The female connectors are on thin backplanes embedded in the 4-cm wide belt. The belt, made of two kapton layers, is adequate to interconnect all the necessary cards. The connection is simply obtained by pushing a card onto a free backplane connector. One of these cards, carrying two 1.5-Volt AA batteries, provides the power to the system. Work by several authors (Friedman et al. 1992, Bedini et al. 1998) helped in the development of the system.

Figure 3. Sketch of the body assembly system.

The main goal of this project was to overcome the usual boxy shape of biomedical devices in order to obtain a true wearability. This multi-sensor monitoring device diverges from the trend of compacting objects and provides an extensive structure, a thin adaptable surface that works as a second skin. The choice of a surface instead of a volume is also based on the goals of properly distributing weight and easily configuring particular settings for different patients. It is a garment that can be worn in different ways: patients can hide it as underwear, body-fitness practitioners can show it off as a fashion object, and athletes can use it as a technical aid.

This monitoring system has been found suitable for both monitoring elderly persons at home and the clinical follow-up of patients after hospital discharge. Two specific applications have been identified:

Monitoring of patients at risk of cardiac diseases ranging from arrhythmic disorders to compensated heart failure. For the latter the integrated information given by ECG and oxygen saturation is particularly important.

Monitoring of motor activities and evaluation of motor capacity in the accomplishment of the motor tasks of daily life. The accelerometers provide global information about the amount of the motor activity performed in a day. This could be a measure of the motor capacity of the subject after proper patient-by-patient calibration. The insoles, in association with the ultrasound inter-foot distance meter, allow us to estimate the instantaneous support surface. This information together with the response of the two channel accelerometer sensors, gives an insight into the reliability of locomotion.

The system has been fully manufactured, but only after a long-term clinical trial will we be able to assess its true potentialities and detect the necessary adjustments to both hardware and software.

Department of Internal Medicine, University of Pisa, Italy

The worsening of cardiorespiratory function is a significant index of moving from compensated cardiomyopathy to heart failure. Starting from the consideration of this harmful pathologic evolution, we set up telemonitoring instrumentation for the early detection of the predictive symptoms of this often fatal condition.

The modular approach, previously described in the design of both the personal device and the central control unit, has led to produce an effective pre-industrial result. Simply combining specific standardized components (analog front-end, processor unit, communication interface), a portable pocket instrument has been obtained capable to log long term EKG and oxygen saturation (more than 24 hours). The instrument is capable of digitally delivering the stored signals through the telephone network (cellular type included) and/or radio and/or IR communication channels. The EKG refers to the standard lead configuration with a dynamic range, frequency response and digital conversion following the European standard for computerized electrocardiography. The oxygen saturation is based on pulse oxymetry and is completely digital, based, assuring complete ratiometric behavior with respect temperature, analog and optoelectronic component characteristics. The control unit modular approach has led to a "Internet server" based design.

Dilated and ischaemic cardiomyopathies represent hard to detect and treat diseases. A major aspect of the cardiomyopathy is a deterioration of both cardiac and respiratory function. The decrease in the contractility of the ventricular walls results in a self-sustained increasing of the heart sizes while the increase in the pulmonary pressure (right heart pressure) results in a markedly depressed respiratory function (Figure 1).

Figure 1. PV loop and atrium pressure varying during dilative myocardiopathy.

The possibility of monitoring, possibly during the normal daily activities, both cardiac function (EKG) and respiratory function (oxygen saturation) can be important when the disease is worsening and during the rehabilitation period, immediately after surgical procedures. In order to accomplish effective monitoring equipment, we developed a dedicated portable instrument, easily linked automatically or on request, to several specialist medical centers even if they are not equipped with special central stations. The modular approach, previously described in the design of both the personal device and the central control unit, has led us to produce an effective pre-industrial result. Work by several authors (Friedman et al 1992, Bendini et al. 1998, Franchi et al. 1998, and Ripoli et al 1998) has assisted in developing this system.

We exploited a basic technology (methods, hardware and software tools) purposely oriented to telemedicine previously developed by the authors. The approach foresees, from the control unit point of view, a new management philosophy, leading to central stations effectively distributed on a network. All the patient devices are wearable. The developed portable system is based on an intelligent and powerful controller provided with large memory capacity. The unit also has a reliable telephone interface for telemetry and telemedicine applications using, if necessary, a digital approach. The signals continuously monitored are EKG and blood oxygen saturation. The devices are controlled by a central station based on a Internet server (Figure 2). For collecting the data, the cardiologist can set up his/her own telemedicine service simply using his/her PC coupled by a modem to the telephone line, provided with a standard Internet browser.

Figure 2. A schematic diagram of the Internet-based control of the personal terminals. The central units can control all the functions of the remote terminals (data, signal retrieval, device pre-setting etc.) by a medical provider, using a standard Internet browser. Any kind of communication means (PSTN, ISDN, etc.) is supported.

To assure the previously mentioned simplicity in using and setting up the necessary telemedicine network, a specialized service, i.e. a true "Cardiologist Medical Provider," supplying the link (hardware, software and data control) for the remote patients is currently being designed.

The modular approach, previously described in the design of both the personal device and the central control unit, has led to produce an effective pre-industrial result. The developed portable unit is schematically represented in Figure 3; it consists of a general purpose system for multifunctional dynamic monitoring of biosignals. The basic version has capability for measurement of blood oxygen saturation and an EKG lead to get the cardiac rhythm. At present, CO₂ monitoring and a proprietary system for continuous non-invasive measurement of blood pressure are also under development. The electronics are equipped with a specialized front-end that basically contains:

An EKG channel (used also to synchronize the other measures).

An optoelectronic system for detecting the peripheral plethysmographic wave (on the finger and/or on the ear lobe); the plethysmographic wave is at present used for the measurement of the oxygen saturation and it will be successively exploited, in conjunction with the EKG signal, for the beat-to-beat evaluation of blood pressure.

Figure 3. A schematic representation and some implementations of the multifunctional ambulatory, tele-controlled terminal.

The arterial blood oxygen saturation is obtained by a digital algorithm, exploiting the blood pulse and the different absorbance of both the total hemoglobin and the oxi-hemoglobin, in the red and near infrared light bands. It is possible, according to the programmed protocol, to record the physiologic variables on all or part of the memory card. The system can also perform front end signal processing, trend production and, finally, digital data transmission to a central control station.

The link to the telephone is provided through a modem, by direct connection to an RS232 part, or as an option, by radio and/ or IR interface (IRDA) as depicted in Figure 4. The use in telemetry nets is accomplished by special embedded communication software and hardware coupling the system to a commercial, lightweight, inexpensive radio-modem module RFM433-LC, produced by the Stamptronic Company, France. Also cellular telephone transmission has been anticipated, using ad hoc developed software on the portable device in conjunction with a GSM cellular apparatus, WM01-G900 by Wavecom company.

Figure 4. Schematic representation of the patient terminal telephone link.

The monitoring system has been tested with eleven patients who underwent major cardiac surgical procedures at the Pasquinucci Hospital (Massa, Italy). The treated cardiopathies were 6 ischaemic and 5 idiopathic. The patients were monitored immediately after their arrival in the recovery room. The monitoring continued after they left the hospital, for a follow-up period ranging from 4 months to 9 months. The information provided by the system was helpful in managing the patients’ rehabilitation.

As demonstrated by the previous description, the present technology appears to be worthwhile to produce inexpensive portable devices for remote patient monitoring using the public telephone lines. The effective diffusion of this methodology is mainly related to two main problems: the clinical use, that up to now has been principally focused on emergency situations, and the organizational engagement of the central control station. These two applications are related; in fact, when the system is devoted to emergency use, continuous medical assistance availability is necessary and also the technical requirements are very stringent. We are also concerned with the use of the telemonitor mainly for the sub-acute, rehabilitative, phase of the illness, in post hospitalization phase and in chronic disease. These applications extend the possible uses of the system and permits the use of existing networks (i.e. Internet backbones) to support a guaranteed "quasi" on line assistance. This moves the technical difficulties to a well established service, the medical provider, that assures all the network services, automatic delivering of special communication/processing software and, finally, the data management. Furthermore, this organization allows the physician to collect data from the remote patient off line and/or on line, by his/her Intranet connection with the local Medical Provider (Figure 4).

The functions implemented up to now in setting-up the pilot Medical Server are summarized as follows:

Managing of the telephone communication with analog (decoding and digital conversion of the signal transmitted as a FM modulated acoustic carrier) and digital (asynchronous up to 19200 baud) techniques;

Automatic computerized ECG analysis, if requested;

Data storing and retrieval;

Communication and application programs and service maintenance with special reference to the data safety and security and external/internal data access.

By means of some free-of-charge telephone numbers, the built up pilot architecture allows communication with the provider; this facility avoids problems to the calling patient, in case of necessity, possibly related to the ISP communication (busy line, temporary unavailability of the ISP-Internet connection, etc.). The access to the medical information by the remote physician PC is managed by sophisticated Internet applications exploiting the Client/Server architecture DCOM (Distributed Component Object Model).

The developed monitoring system offers, in our opinion, an important aid in the management of the rehabilitative phase of surgically treated heart failure. Even if a clear understanding of the exact mechanisms that cause the further deterioration of the disease has not been reached, surely the continuous acquisition of clinical parameters related to cardiac and respiratory function, during everyday activity may constitute a valuable base of knowledge for surgeons and physicians, resulting in a safer rehabilitation for the patients. In fact, the continuous monitoring of both ECG and respiratory parameters (i.e. Sa0₂ and C0₂) gives important information about the pulmonary pressure and the related ECG trends. As a matter of fact, the increase in pulmonary pressure results in an immediate decrease of blood oxygen saturation, in such a way that the device offers an indirect measure of the right heart pressure, and the related ECG wave integrates the information about the whole cardiorespiratory response.

Clinica Medica IV, Dipartimento di Medicina Clinica e Sperimentale,
Università di Padova, Via Giustiniani 2, 35128 Padova, Italy

Reparto di Cardiologia, Ospedale di San Daniele del Friuli, Udine, Italy

Clinica Medica IV, Dipartimento di Medicina Clinica e Sperimentale,
Università di Padova, Via Giustiniani 2, 35128 Padova, Italy

Centro di Cardioreumatologia, Ospedale di Pordenone, Pordenone, Italy

Divisione di Medicina, Ospedale di Dolo, Venezia, Italy

The aim of the study was to assess the effect of physical activity on 24-h ambulatory blood pressure (ABPM) and office blood pressure (BP) in 1,091 subjects with borderline to mild hypertension. Subjects were 18-45 years old with diastolic BP (DBP) from 90-99 mmHg and/or 140-159 mmHg who never took antihypertensive medication. All subjects underwent physical examination, office BP measurements, and two 24-h ABPM three months apart. Subjects were classified as non-exercisers, Group 1 (451M:241F), mild exercisers, Group 2 (260M:63F) and heavy exercisers, Group 3 (74M:2F). During the three-month follow-up, subjects maintained the same physical activity habits. There was no difference in smoking and alcohol consumption between the 3 groups. Because the groups differed significantly in age (34±0.3 vs. 31±0.5 vs. 29±1 yr), body mass index (25.8±0.1 vs. 24.8±0.2 vs. 24.5±0.1) and gender data were adjusted for these confounders. Data and changes in BPs, between baseline and the third month were compared with a Student’s t-test (data expressed as MEAN±SEM). Office systolic BP (SBP) was higher in Group 3. Systolic 24-h ABPM was similar in the three groups (24-h 130.6±0.4 vs. 130.8±0.6 vs. 132.2±1.3, mmHg; p=ns) and the same trend was observed in daytime and nighttime BP. Diastolic 24-h ABPM was inversely related to the intensity of exercise (82.6±0.3 vs. 80.1±0.4 vs. 77.4±0.9, Group 1 vs. Group 2 p<0.09, Group 1 vs. Group 3 p=0.001) as well as daytime and night time. HR was always inversely related to the exercise intensity. There were no changes in office SBP and HR between baseline and third month follow-up while office DBP slightly decreased in the exercisers (Group 1 vs. Group 3 p=0.05, Group 2 vs. Group 3 p=0.03). The difference between systolic ABPMs at the third month and baseline decreased significantly according to exercise intensity (24-h Group 1 vs. Group 3 p=0.005, Group 2 vs. Group 3 p=0.007; daytime Group 1 vs. Group 3 p=0.004, Group 2 vs. Group 3 p=0.007; nighttime Group 1 vs. Group 3 p=0.02, Group 2 vs. Group 3 p=0.04). No changes in diastolic ABPMs were measured. In conclusion, physical activity has a positive effect in lowering BP attenuating the risk of hypertension in young subjects with borderline hypertension. The decrease in BP is better detected by ABPM and this effect is still present after three months of exercise, and the group with sustained physical activity has an additional reduction in SBP according to ABPM detection.

Several studies have shown that subjects with borderline to mild hypertension are at increased risk of morbidity and mortality (Morris et al. 1980, Paffembarger et al. 1986). According to international guidelines the decision on whether to treat young subjects with mild hypertension should be based on office BP, ambulatory BP and target organ involvement (JNC VI). The benefit of physical training in reducing the rate of cardiovascular disease through the reduction in blood pressure (BP) and other risk factors is well established (Hagberg 1990). For this reason exercise is now recommended in the management of hypertension in the early phase of the disease or in association with pharmacological therapy. Much less is known on whether the antihypertensive effect of physical activity is constant in time and whether there is any difference between traditional office measurement and 24-h ambulatory BP measurement

Seventeen hospital departments in northeastern Italy participate in the HARVEST Study (Palatini 1993). This study was directed at evaluating the predictive value of ambulatory BP monitoring for the development of fixed hypertension in patients with borderline to mild hypertension. One thousand ninety-one subjects aged 18 years to 45 years with diastolic BP ranging from 90 mmHg to 99 mmHg and/or systolic BP between 140 mm Hg and 159 mmHg, who never took any antihypertensive therapy were evaluated. The protocol was approved by the Ethics Committee of the trial (Palatini 1993) and the subjects gave informed consent.

All subjects had a complete medical history taken and an interview about their physical activity habits, smoking, alcohol intake and coffee. All subjects underwent physical examination, anthropometry, blood chemistry analysis, office and two 24-h ABPM taken three months apart. Echocardiography was done on 744 subjects, and they had 24-hour urine collection for catecholamine determination (Rossi et al. 1986).

Subjects were classified according to the intensity of physical exercise according to: 1) fully sedentary, 2) mild exercisers if they practiced mild level of physical activity, 3) heavy exercisers if they exercised at least 3 times per week. Neither duration nor intensity level of activities was considered in the classification to keep physical activity assessment as objective and free of bias as possible. Subjects were defined as non-smokers if they did not smoke or smokers if they smoked one or more cigarettes per day. Coffee consumption was defined according to the number of cups of coffee drunk per day.

BP was measured according to the recommendations of the British Society of Hypertension (British Hypertension Society 1989). The mean of six readings taken in the supine position during two separate visits performed 2 weeks apart was defined as office BP, except in subjects with sounds tending to zero, in whom Phase IV was taken.

In 88% of subjects, 24-h AMBP recordings were obtained with the A&D TM-2420 Model 7 which uses a microphone to detect Korotkoff sounds, and in the other 12% used the ICR Spacelabs 90207, which employs the oscillometric method. Both of these devices were previously validated (Palatini 1994). Before starting the study, all investigators underwent common training procedures on the use of the recorders to ensure methodological homogeneity. During the recordings, subjects were invited to follow their ordinary routine. In the exercisers, ambulatory monitoring was performed on a non-active day. Subjects were asked to go to bed no later than 11 pm. BP was measured every 10 min during waking time (6 am to 11 pm) and every 30 during the night.

M-mode and 2D echocardiography was obtained in 744 subjects. Left ventricular internal diameter was measured at end diastole, according to the American Society of Echocardiography (Sahn et al. 1978). Left ventricular mass was calculated according to the following formula: 0.8(1.04(IVS+LVDD+PWT)³-LVDD³)+0.6g (Devereux et al. 1986), where IVS is interventricular septum thickness in diastole; PWT is posterior wall thickness in diastole; LVDD is left ventricular end-diastole. Left ventricular mass was normalized for body surface area. Left ventricular wall thickness was defined as the sum of interventricular septum thickness in diastole + posterior wall thickness in diastole and relative wall thickness as the ratio of wall thickness to end-diameter in diastole. Left ventricular filling rate was assessed by Doppler analysis of transmitral flow, according to a procedure described elsewhere (Bongiovi’et al. 1992). Maximal velocity of early (Evmax) and atrial (Avmax) waves were measured, and the EV max/AV max ratio was calculated. All measurements were made by the same experienced physicians.

Statistical comparison between means was assessed by ANOVA, and c 2 analysis was used for categorical variables. As the three groups differed in age, BMI and gender, the differences were adjusted for those confounders. The difference between baseline and third month was calculated for BP measurements and data were analyzed by a paired Student’s t-test. Data are expressed as mean ± SEM unless specified. A p<0.05 was considered statistically significant. The SAS statistical program was used for statistical analysis (SAS, Inc., Cary, North Carolina).

There were 692 subjects (451 men and 241 women) categorized as non-exercisers (Group 1), 323 subjects (260 men and 63 women) classified as mild exercisers (Group 2), and 76 subjects (74 men and 2 women) classified as heavy exercisers (Group 3). There were no differences in smoking habits and alcohol intake between the groups.

Age decreased according to exercise intensity. Non-exercisers were 34.4±0.3 years old, mild exercisers were 31.8±0.5 years old and the heavy exercisers were 28.7±1 years old (Group 1 vs. Group 2 p<0.0001, Group 1 vs. Group 3 p<0.0001, Group 2 vs. Group 3 p=0.005). Serum cholesterol (Group 1: 201.6±1.6 mg/dl, Group 2: 195.6±2.4 mg/dl, Group 3: 190.6±4.8 mg/dl; p=ns) and triglycerides (Group 1: 113±2.7 mg/dl, Group 2: 111.4±4.1 mg/dl, Group 3: 105.1±8.4 mg/dl; p=ns) were similar after adjustment for age, gender and body mass index. Body mass index was higher in the non-exercisers and similar in the other two groups (25.8±0.1, 24.8±0.2, 24.5±0.4; Group 1 vs. Group 2 p<0.0001, Group 1 vs. Group 3 p=0.002, Group 2 vs. Group 3 p=ns). Weight was constant during the period of follow-up (baseline: Group 1: 77±0.8 Kg, Group 2: 76.4±1.1 Kg and Group 3: 80.5±2; Group 1 vs. Group 2 p=0.001 and Group 1 vs. Group 3 p<0.05; third month: Group 1: 75.2±0.5 Kg, Group 2: 76.2±0.8 Kg, Group 3: 80.5±1 Kg, Group 1 vs. Group 3 p=ns) and also the difference between baseline and follow-up was not statistically significant.

There were no differences in the echocardiographic dimensional indexes among the three groups although the LV mass increased as physical activity increased but the difference did not achieve the level of statistical significance. LV diameter in diastole was significantly greater in the third group. The EV max/AV max ratio was slightly greater in the exercisers, but the difference was not significant also after adjustment for heart rate (Table 1). Twenty-four hour urinary norepinephrine/creatinine (59.9±3.0 Group 1, 55.1±4.5 Group 2, 50.6±9.8 Group 3) and epinephrine (23.6±0.8 Group 1, 27.1±2.7 Group 2, 23.3±5.6 Group 3) were not different among the three groups.

Table 1. Echocardiographic parameters.

Heart rate either measured in the office or by ambulatory BP monitoring was inversely related to exercise intensity. At baseline office systolic BP was significantly higher in the group of heavy exercisers and office diastolic BP was lower. There were no differences in systolic ambulatory BP recordings while diastolic BP decreased as the intensity of exercise increased. After three months only office diastolic BP was significantly lower in the exercisers compared to the non-exercisers while the other BPs were similar among the three groups (Table 2). When the difference between BPs measured at the third month and BPs at baseline were analyzed, the group who practiced heavy sport activity showed and additional significant decrease in systolic BP detected by ambulatory BP monitoring (Figure 1). Although heart rate decreased by 2 beats per minute in the third group, after three months’ exercise, the difference was not statistically significant.

Figure 1. Difference between systolic blood pressure at the third month and systolic blood pressure at baseline.

Table 2. Office and ambulatory blood pressure at baseline and after 3 months.

The relationship between physical activity and BP was intensively investigated by longitudinal and cross-sectional studies, either in normotensives or hypertensives and, with few exceptions, it has been shown that physical activity is beneficial in lowering BP levels. We have shown, in a previous paper, that ambulatory BP monitoring could be lower than office BP in those subjects who perform regular exercise but there are few other reports on the time-course effect of exercise on ambulatory BP monitoring (Palatini 1998).

At baseline, office systolic BP was significantly higher in the exercisers but similar in the three groups when measured by 24-hour recording. These results are of clinical significance because they might influence the decision on whether to treat or not borderline hypertensive subjects. Diastolic BP was lower at baseline in the group of heavy exercisers, either measured with the traditional method or by automatic recording. As reported by us and other authors (Palatini 1998, Lehmann et al. 1992), exercise training might decrease BP, at baseline, by reducing the BP reactivity to the stress of daily activities mediated by the sympathetic nervous system. In fact, the lower levels of norepinephrine, also not statistically significant, and heart rate found in the active subjects suggest decreased activity of the sympathetic nervous system. This would explain why daytime diastolic BP was mainly affected instead of night time diastolic BP.

After three months of exercise there was a decrease in systolic BP either at office measurements or by automatic recordings, but the decrease in ambulatory systolic BP, calculated as the difference between systolic BP at the third month and the baseline, was statistically significant in the group of heavy exercisers after adjustment for age, gender and body mass index. Douglas et al. (1991) reported a decrease in ambulatory systolic BP after 12 months of exercise, but their population and the type of exercise were quite different from ours. This means that the effect of physical activity has an additional affect on the regression to mean and that this effect is independent of gender or body mass index. Office diastolic BP was statistically lower in the exercisers but these results were not confirmed by ambulatory BP monitoring. The fact that the group of non-exercisers had a slight decrease in diastolic BP, sufficient to make the original significant difference disappear, could be explained by a change of lifestyle. Probably this group of hypertensives engaged in some sort of physical activity after the baseline visit.

But do casual measurements reflect training-induced changes in arterial BP throughout an entire day after three months of exercise? These results demonstrate that casual BP and 24-h ambulatory recording can provide quite different information regarding the influence of physical activity on chronic levels of arterial BP. Several laboratories have reported reduction in casually recorded BP levels in non-exercisers similar to exercisers which could be due to an insufficient period of observation. This was also true in our study but 24-hour recording could detect an additional, independent effect of physical activity. Three months of training were associated with a 2-beat reduction in heart rate in the group of exercisers which might explain the decrease in 24-h systolic BP. The slight but significant reduction in 24-hour systolic BP could contribute to protect target organs such as the myocardium, which better correlates with systolic BP than diastolic BP and the kidney.

There were no differences in LV mass index among the three groups. The lower ambulatory BP values found in the exercisers could counterbalance the trophic effect of training and thus explain why LV mass in hypertensive trained individuals is not greater than in sedentary subjects with similar office BP. At variance with other studies, we found no differences among the three groups in LV diastolic function. In a previous study of ours we reported that LV structure involvement is the earliest sign in hypertension, and as our subjects had similar LV structure, it is conceivably why they had also similar diastolic function. Probably, physical activity per se, does not improve diastolic function as in normotensive subjects (Palatini 1998).

In conclusion, physical activity has a positive effect in lowering BP attenuating the risk of hypertension in young subjects with borderline hypertension. The antihypertensive effect of physical activity persisted after three months and the group of exercisers had an additional reduction in systolic BP detected by ambulatory BP monitoring. To obtain accurate information on chronic levels of arterial pressure over time, 24-hour ambulatory BP should be performed whenever possible along with traditional casual readings.