In home telerehabilitation for geriatric patients


Published on

  • Be the first to comment

  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

In home telerehabilitation for geriatric patients

  1. 1. GERONTECHNOLOGY In-Home Telerehabilitation for Geriatric Patients© DIGITAL STOCK & EYEWIRE Use of Wearable Wireless Body Area Sensor Networks for Increased TelepresenceBY MATHIEU HAMEL, REJEAN FONTAINE,AND PATRICK BOISSYI n the last decade, changes in the organization and financing dedicated modular software interfaces for user-friendly control of health services in Canada have reduced the length of stay of videoconferencing connections, PTZ camera function, and in acute care hospitals, increased the number of day sur- external devices (i.e., tablet PC and sensors). An overview of geries, and generally reoriented the hospital-centric care the telerehabilitation platform and the software interface for thetoward dispensation of health services in the community. The clinician is illustrated in Figure 1.demographic imperative of an aging population creates unique Iterative changes were made to the hardware and softwareopportunities to look at new paradigms in delivering health components to ensure transparent dynamic interactions be-care services in the community. In this context, in-home telere- tween the clinicians and the clients during a telerehabilitationhabilitation (i.e., the delivery of rehabilitation services at an session. Special attention was given to provide a mouse-basedindividual’s home over telecommunication networks) has been interface to control intuitively from a unique screen throughidentified as a promising avenue. The rationale for in-home tel- point-and-click or area-zoom PTZ camera functions at botherehabilitation is to expand and facilitate the delivery of reha- sites. Results from our ongoing trial and debriefing of cliniciansbilitation services to people who cannot travel to a clinic have shown that telerehabilitation practices challenge conven-because of disability or travel time [1], [2]. Evidence support- tional communication behaviors underlying the professionaling the use of telerehabilitation as a viable alternative or com- patient-client relationship found in face-to-face encounters inplement to traditional in-home therapy is slowly emerging in rehabilitation. Although videoconferencing can create a tele-the literature [3], [4]. presence experience for the clinician by providing visible and Most types of telerehabilitation services fall into two catego- nonverbal information about the behavior of an individual inries: clinical assessment (the patient’s functional abilities in his/ his/her environment, it is difficult for the clinician to interprether environment) and clinical therapy. To provide both types of detailed information such as the kinematics and kinetics of theservices remotely while interacting with the patient, the rehabil- individual’s movement and physiological responses to exer-itation professionals rely on establishing a telepresence through cises in a telerehabilitation context. This is even more evidentbidirectional video and audio from videoconferencing equip- when operating under suboptimal optical conditions such asment connected through a high-speed Internet connection. those found in the home environment. Increased telepresenceTelepresence [5] refers to the phenomenon whereby a human combining information from wearable sensors with audio andoperator develops a sense of being physically present at a video streams might be part of the solution to complement theremote location through interaction with the user and the subse- traditional telerehabilitation practices [9], [10].quent perceptual feedback he/she receives via the appropriateteleoperation technology [6]. Wireless Body Area Sensor Networks We investigate in this study, following the positive results Wireless body area sensor networks (WBANs) are well suitedfrom a proof-of-concept study [7], the effectiveness of provid- to increase telepresence, as they can provide specific informa-ing in-home telerehabilitation services as an alternative to tion about an individual’s behavior without using complexhome care visits for physical therapy in orthopedic conditions laboratory equipment and without interfering with the person’sfollowing discharge from an acute care hospital and rehabilita- natural behavior [11]. WBANs are generally built aroundtion unit [8]. Based on the results from the initial proof- several sensing devices wirelessly linked together using nar-of-concept study and a user-centered design approach, a row-band radio communication [12]. Recent developments intelerehabilitation platform was developed consisting of two the field of wireless networks have generated many newH264 videoconferencing codecs (Tandberg 500 MXP) with commercial wireless communication platforms based on dif-integrated wide-angle view cameras and remotely controlled ferent protocols and technologies (Wi-Fi, WiMax, Bluetooth,pan tilt zoom (PTZ) functions, local and remote computers with Zigbee, UMTS, UWB) [13]. These technologies offer a wide range of characteristics in terms of speed, transmission range,Digital Object Identifier 10.1109/MEMB.2008.919491 power requirements, connectivity, and cost. The choice ofIEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE 0739-5175/08/$25.00©2008IEEE JULY/AUGUST 2008 29
  2. 2. wireless network architecture Clinical Site for a WBAN application is 2 context and sensor depend- ent. Table 1 presents some Home Site of the existing BAN/WBAN 1 technologies and their wire- less networking character- istics. The use of a WBAN system in a telerehabilitation context calls for a small, reli- able, low-power platform ca- pable of seamlessly integrating several modules. The Zigbee technology was designed for this type of appli- cation. The IEEE 802.15.4 physical radio standard oper- (a) ates on the 2.4-GHz unli- censed band over 16 channels, and the network layer supports topologies such as star, tree, 1 and mesh. Depending on the power output and environ- mental characteristics, trans- mission distances range from 10–100 m [14]. Recent publi- cations [11], [15], [16] have 3 illustrated projects geared to- ward developing application- specific WBAN systems based on Zigbee technologies. Recommendations on a mul- 2 titier architecture for WBAN systems in the context of patient monitoring or the types of sensors to use and their lo- cations have been proposed [15], and different WBAN systems are currently under development. ActiS, an ac- tivity sensor developed by Jovanov, is built around a (b) wireless platform that integra- tes a Zigbee-compliant radio Fig. 1. Telerehabilitation platform. (a) Hardware components including two H264 videocon- and a microcontroller called ferencing codecs (Tandberg 500 MXP) with integrated wide-angle view cameras and Telos from Moteiv [17]. A remotely controlled PTZ functions. (b) Software interface for user-friendly control of video- custom sensor board con- conferencing connections, PTZ cameras function, and external devices (i.e., tablet PC and nected to the Telos platform sensors). enables concurrent wireless ECG and accelerometer mea- surements. As a heart sensor, ActiS can be used to monitor Table 1. Wireless technologies and possible BAN/WBAN platforms. the heart activity and trunk position. CodeBlue is another Technology Transfer Rate Range BAN/WBAN project developing wireless Wi-Fi 11–54 Mb/s 30–50 m DPAC Airborne, PDAs body area networks for medi- WiMax 4.5–70 Mb/s 100 m–50 km Portable computers cal care. The goal of the Bluetooth 57 kb/s–3 Mb/s 100 m Smart-Its, iMotes project is to develop sensors Zigbee 20–250 kb/s 100 m MICAz, Telos, tMotes for stroke rehabilitation patients UMTS 50 kb/s–2 Mb/s 5–100 km Mobihealth and to monitor vital signs to UWB 54 kb/s–48 Mb/s 1–10 m Magnet help in emergency response (ECG, blood pressure) [18].30 IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE JULY/AUGUST 2008
  3. 3. WBAN Home Site 1 3 5 2 4 1 WBAN Transceivers 2 WBAN Receiver 3 Tablet PC 4 Tandberg 550 MXP 6 Clinician Site Pulse 9 Oximeter + 8 Accelerometers Instrumented Soles Respiratory Belt Sensors 7 5 VPN Router 8 Tandberg 550 MXP 6 Internet 9 Clinician PC 7 VPN Router Fig. 2. Complete system used for a telerehabilitation session. The WBAN system comprises four wireless sensor nodes. A total of 32 analog signals are sampled at 100 Hz frequency and sent to the host computer. Sensors measure the heart rate, blood saturation, changes in thoracic and abdominal circumference, weight-bearing, acceleration, and angular rate. Video, audio, and sensor data are sent to a remote site using a high-speed Internet connection.The wireless platform chosen for this project is the MICAzfrom Crossbow [19], which is also based on a Zigbee- Transmitter Receiver Flash Flashcompliant radio. Memory MemoryWBANs for Telerehabilitation Processor Processor Sensors Analog I/O Analog I/O PC Digital I/O Digital I/OSystem Architecture 2.4 GHz 2.4 GHzFor use in telerehabilitation applications, we recently developed Radio Radioa Zigbee-based WBAN system with custom sensor platformsand adaptable sensing inputs capable of accommodating differ- Wirelessent sensor configurations. The system designed for telerehabi- Modulelitation applications is composed of sensor platforms withapplication-specific signal conditioning units connected to wire-less communication modules. An overview of the system archi-tecture and components is illustrated in Figure 2. The systemconsists of four eight-channel Zigbee-based wireless sensornodes with a total theoretical bandwidth of 250 kbps configured Li-Ion Sensor Board Batteryin a star configuration to a single receiver connected to acomputer. The current sensor node configuration comprisesa custom sensor board with an embedded three-dimensional Fig. 3. WBAN and sensors. Wireless sensor network comprisesaccelerometer (LIS3L02AQ, STMicroelectronics) [20], one one- up to four sensor nodes configured with the star topology.dimensional gyroscope (ENC-03M, Murata) [21], and connectiv- Wireless modules include a custom sensor board and aity to four external analog or digital sensors (Figure 3). External MICAz communication module from Crossbow Technology.sensors can take many forms: we currently use load cells, respira-tory belts, and a pulse oximeter. The two respiratory belt sensors external sensors described in this article (oximeter, respiratory(MLT1132, ADInstruments) [22] are connected to the first sen- belts, and the instrumented shoes) can all be installed with nosor node worn on the trunk. The second and third sensor nodes or minimal exterior help. The modules, as shown in Figure 2,are linked to custom instrumented shoes, which provide weight- have elastic bands and adjustable bracelets that enable the sub-bearing data during ambulatory activities. The last sensor node jects to install them with relative ease. In certain cases, individ-uses onboard sensors to measure acceleration and angular rate of uals with reduced mobility or dexterity (e.g., stroke) could getthe subject’s dominant hand. assistance from a third party to install the sensor module if In the context of telerehabilitation, sensor placement is a needed.critical issue. While the ergonomics, usability, and design of The communication module is an off-the-shelf MICAzwearable sensors can affect the reliability of the data, the available from Crossbow [19]. The module consists of anIEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE JULY/AUGUST 2008 31
  4. 4. ATmega128 microcontroller with eight 10-b analog-to-digital to the videoconferencing equipment (Tandberg 550 MXP, converters (ADCs), flash memory, and a Chipcon 2.4-GHz H.264 codec). A secured VPN communication channel was radio transmitter/receiver. Modules can be programmed as established between the two sites using a second identical receivers, transmitters, or both using an event-driven, highly router at the clinical site. Raw signals provided by the wireless modular operating system called TinyOS [23]. This operating sensors (Figure 4) can be directly visualized at the clinical site system is based on a library of components that can be easily and further be processed through an algorithm that interprets connected using well-defined interfaces. Custom components, in real time the variables such as body angles, weight-bearing, written with the NesC language [24], can directly interact with respiration, and heart rates [Figure 1(b)]. components from the TinyOS library with minimal use To assess the feasibility of using the proposed WBAN sys- of resources. The network is formed by assigning a unique tem with the existing telerehabilitation platform, we evaluated address to each wireless module individually. The main its radio communication performance, operational range, and receiver module acts as a coordinator by sending start and stop functionality under telerehabilitation conditions. More specif- commands to transmitters, enabling synchronized data acquisi- ically, the objectives of the system’s evaluation were to 1) tion. Small 580-mAh Li-ion batteries (UBP363450/PCM) assess the impact of the number of sensor nodes used, the power both the sensor boards and the communication modules number of sensor inputs per node used, and the sampling rate and are embedded in bracelets that can be attached to the body. used on the reliability of the radio communication; 2) charac- The WBAN is configured with four wireless sensor nodes. terize the performance of this system during continuous use in A tablet PC served as the WBAN receiver at the home site and a home environment; and 3) assess the performance in con- was connected wirelessly (802.11 b) to a router [Linksys with junction with a videoconference link over the Internet. virtual private network (VPN)] connected to a digital subscriber line (DSL) modem for Internet access. The WBAN System Evaluation was connected to the computer via a USB interface board Although the Zigbee-based WBAN systems described in (MIB520). The router also provided wireless Internet access the literature are quite innovative and their development is A A Left Foot 0–900 N Right Foot Weight Bearing B ±2G aX aZ WY Left Ankle aY Right Ankle Left Wrist Abdominal Respiratory Thoracic Belt Transducer 97% 87 bpm % Oxygen Saturation 86 bpm % Pulse Rate 96% 96% 85 bpm Pulse Oximeter 1s 2s 3s Fig. 4. Signals output from the WBAN system during a walking activity. The A and B cycle shows the applied vertical forces on the insoles and the leg movements during a normal walk cycle. Activity levels can be cal- culated by combining heart rate, respiratory data, and a sum vector of accelerometer signals.32 IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE JULY/AUGUST 2008
  5. 5. ongoing, there are little or no published data concerning the data packet was used for these experiments. Transmissionperformance and limitations of these systems in terms of radio errors are either a missing message from one of the transmitterscommunication, operational range, and functionality under or a message not received in the right order. During the experi-unconstrained conditions in a home environment when worn ments, precautions were taken to make sure that the batteriesby an individual. Indeed, proximity issues regarding the place- of each module were properly charged and that the modulesment of several modules on the body may lead to severe inter- were worn correctly (two bracelets on the wrists and two on theference problems, and the reliability of continuously streaming leg shank). The distance from the receiver (DFR) was alsohigh volumes of data to a receiver at a determined rate over a standardized between tests, making sure it would not exceedlong period is untested. 50 ft between the receiver and the person. A total of 25 trials of Wireless platforms such as Moteiv, MicaZ, and other Zigbee- 30 min each were performed while varying the number of mod-compliant devices were mainly developed for commercial and ules from one to five and the sampling frequency from 50 toindustrial practices. The goal of the following experiments was 100, 200, 400, or 800 Hz. During each trial, tasks related toto establish the performance of a typical WBAN in real condi- office work were done (walking, typing on the computer, etc.).tions to provide guidelines for future WBAN development and Results illustrated in both graphs of Figure 5 summarize theimplementation. First, a reliability experiment was conducted to performances of several WBAN setups and suggest a typicaldetermine the performance of several WBAN configurations in network comprising four active modules, which minimizes thean ideal laboratory environment. Second, a similar experiment probability of communication errors and optimizes the numberwas done in a home environment to evaluate the effect of this of active modules and their sampling rates.environment on the WBAN system. Finally, the last experimentconsisted of streaming data from the WBAN system in the con- Real Home Environmenttext of in-home telerehabilitation (i.e., shared bandwidth be- Special precautions have to be taken considering that thetween videoconferencing equipment and the WBAN system WBAN system would be used in a home environment.over a DSL Internet connection). Although laboratory experiments give an idea of system performance, it is essential to evaluate the system in a typicalWBAN Reliability in a Laboratory Environment home with interwall and interfloor communication, possibleThe purpose of the reliability experiment was to determine the sources of signal reflections and noise. The purpose of thisbandwidth limitations of this kind of system in a controlled experiment was to determine whether or not it is viable to uselaboratory environment and evaluate the possible problems this system in an in-home telerehabilitation context. Assump-related to interference and body movement. Several tests were tions were made concerning the communication algorithmconducted while varying the number of active modules and and bandwidth requirements, as these parameters could bethe sampling frequency of the ADCs. The system reliability optimized. The system configuration used during these testswas evaluated by assuming that communication errors would was taken from previous results derived from the laboratoryhappen independently of the algorithm programmed in the experiments (four modules, 100 Hz). A typical multilevelmicrocontrollers. It is also possible to avoid transmissions house was chosen as the testing environment (Figure 6).errors by programming a more robust error detection algorithm The receiver module and host computer were situated onthat would send back bad or missing data packets. A simpler the second floor of the house. Two parameters were eval-algorithm that associates a message number and origin of each uated during the trials: the percentage of communication 30 100% Time (min) to Loss of Communication 25 1 Module 1 Module 2 Modules 2 Modules 3 Modules 3 Modules % Transmission Errors 4 Modules 0.7 4 Modules 20 5 Modules 5 Modules 0.6 15 0.5 0.4 10 0.3 0.2 5 0.1 50 Hz 100 Hz 200 Hz 400 Hz 800 Hz 50 Hz 100 Hz 200 Hz 400 Hz 800 Hz (a) (b)Fig. 5. WBAN reliability experiment results. (a) Total time required over a 30-min experiment before losing communicationbetween the WBAN transmitters and the receiver for multiple setups. (b) Percentage of transmission errors during these experi-ments while varying the number of modules and the sampling frequency of the eight analog inputs.IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE JULY/AUGUST 2008 33
  6. 6. errors as described earlier in this article and the number of appliances, and body movements did not prevent the system communication losses. This last parameter was evaluated by from working properly. counting the number of times the system completely lost track of the wireless network during the 1-h trials. The vari- WBAN in a Telerehabilitation Context ous activities performed during these trials included both A last experiment was conducted to evaluate the performance static and dynamic movements. More precisely, activities in and impact of the WBAN system when used in conjunction Room 1 (office) included writing in a sitting position, vac- with a videoconferencing system (i.e., shared bandwidth). A uuming, and tidying the closet. In Room 2 (second-floor bed- 60-min telerehabilitation session took place at the home site, room), vacuuming and office work were done. Cooking and data from the WBAN were sent to the host computer (clini- lunch, washing the dishes, and cleaning were done in Room cal site) in real time via a high-speed Internet connection. Fig- 3 (kitchen). Activities in Room 4 (dining room) included ure 2 shows the system used during this experiment. A DSL vacuuming and computer work (sitting at the dinner table). high-speed Internet access at the two sites provides a theoreti- Room 5 (bathroom) included some laundry, vacuuming, and cal bandwidth of 3 Mb/s in download and 800 kb/s in upload. scrubbing. Finally, some vacuuming and reading (lying on From this available bandwidth, 384 kb/s was dedicated to the the bed) were done in Room 7 (first-floor bedroom). The videoconferencing equipment to establish a quality audio and mean linear distances between the receiver and the WBAN video link (320 kb/s for video data and 64 kb/s for audio data). transmitters (DFR) were computed using the mid-point Bandwidth allocation was estimated experimentally during (length, width, height) of each room and the receiver location the telerehabilitation session using communication statistics on the second floor. The WBAN performances obtained in (upload and download transfer speed) computed by the router each room in terms of communication errors and loss of com- and the videoconferencing equipment located at the clinical munication did not differ from the performances obtained in site. Communication statistics were retrieved from both devi- the laboratory environment. The effect of walls, electrical ces at 5-s intervals. Bandwidth allocation for the WBAN was calculated as the bandwidth statistics recorded on the router minus the bandwidth Second Floor statistics provided by the Area: 14.33 m 2 2 videoconferencing unit. DFR: 6.87 m Continuously polling the sta- Receiver tistics also requires part of LOC %ERR the total bandwidth for both 1 0.098 upload and download. It was included in the WBAN band- Area: 15.68 m2 1 width for simplicity. Results DFR: 1.68 m from this experiment are il- lustrated in Figure 7. LOC %ERR Area: 15.65 m2 6 The WBAN setup used for 0 0.158 DFR: 6.93 m the experiment accounted for LOC %ERR approximately, on average First Floor 2 0.151 over the 60-min session, 209 kb/s of the total band- Area: 15.02 m2 7 width allocated during the ses- DFR: 1.26 m sion. The bandwidth needed to stream the data from the LOC %ERR WBAN system in real time 0 0.134 Area: 10.33 m2 4 DFR: 7.86 m over the Internet did not affect the overall quality of the audio Area: 6.39 m2 5 LOC %ERR and video signals received DFR: 3.51 m 3 0.128 from the videoconferencing equipment. During the experi- LOC %ERR ment, the WBAN encountered 2 0.123 communication errors and was restarted three times. The Area: 12.94 m 2 3 Tandberg unit recorded 192 Area: Total Area of the Room (m2) (download) and 95 (upload) DFR: 5.12 m DFR: Mean Linear Distance from Receiver (m) missing data packets through- LOC: Number of Loss of Communication LOC %ERR out the session. %ERR: Percentage of Communication Errors 1 0.082 Power Consumption A custom sensor board Fig. 6. In-home communication reliability results. Percentage of transmission errors and number was built for the telerehabilita- of communication losses during a 1-h continuous transmission from all rooms in a typical house tion application. As described using four modules at a 100-Hz sampling rate (eight channels). in the ‘‘System Architecture’’34 IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE JULY/AUGUST 2008
  7. 7. Download Upload 700 500 Total BW (588 kb/s) Total BW (407 kb/s) 600 400 Bandwidth (kb/s) 500 Bandwidth (kb/s) 400 300 Video (315 kb/s) Video (228 kb/s) 300 200 200 WBAN (209 kb/s) 100 Audio (64 kb/s) 100 0 Audio (64 kb/s) 0 0 30 60 0 30 60 Time (min) Time (min) (a) (b)Fig. 7. WBAN performances and bandwidth allocation during a 60-min telerehabilitation session. (a) Bandwidth allocationduring download composed of video, audio (Tandberg 550 MXP), and sensor data from the WBAN. (b) Bandwidth allocationduring upload composed of audio and video only.section, it contains a tree axial accelerometer, a gyroscope, Table 2. Theoretical power consumption of WBANand four amplifiers for external circuitry. The theoretical sensor nodes.power consumption is presented in Table 2. The total batterylife was tested experimentally during the real home environ- Active Items Operating Currentment experiment. Continuous transmission of the WBAN Wireless module (Crossbow) 17 mA* (Tx mode)lasted until the first module had no power left. A total battery Accelerometer (STElectronics) 1.5 mAlife of 24 h was expected based on the theoretical power con- Gyroscope (Murata) 5 mAsumption (Table 2). Operating current from onboard sensors Amplifiers (Analog Devices) 0.250 mA(accelerometers, gyroscopes, and amplifiers) was added to the Total 23.75 mApower consumption of the MICAz module in continuous Battery life (Li-Ion at 580 mAh) 24.42 htransmit mode. Experimental results suggested a total batterylife of approximately 15.45 h.Discussion Onboard data processing can be achieved to substantially reduce the overall dataflow by transmitting the already ana-Performance and Limitations of In-Home WBAN lyzed data and warnings to the clinician, as suggested in otherDuring the reliability experiments, the system’s farthest limits studies [11], [12], [15]. Event management, as described byof radio communication were tested in terms of sampling fre- Otto, would considerably reduce the overall transmit rate byquencies and number of active sensor nodes. From the results recognizing characteristic features of raw sensor data. How-shown in Figures 5 and 6, it is possible to determine some sort ever, onboard data processing also has a great impact on powerof comfort zone where the proposed WBAN system works consumption and signal latency. Compared with long-timewell and minimizes the probability of errors. This information monitoring scenarios, telerehabilitation sessions are relativelygives a starting point for using a WBAN system during telere- short (1–2 h) and require real-time data transfer for quick accesshabilitation sessions. An optimal configuration consisting of by clinicians. A compromise solution between the amount offour active sensor nodes, each capable of accommodating computing done and overall bandwidth usage should be consid-eight sensor inputs and a sampling rate of 100 Hz, was found ered. No consensus has yet been reached in regard to choosingto offer the most reliability. It should be noted that this particu- the right sensors and data format relevant to clinicians. Thelar setup is a compromise solution between the number of multitude of applications makes it difficult to obtain a uniqueactive modules and the bandwidth requirement of the body standard. During experiments, signals from onboard sensorssensors. Results also showed the possibility of using five were transmitted in their raw format, as this gives importantactive modules by using a 50-Hz sampling frequency or by information about the limitations of the system.taking only four of the eight available analog inputs. The sys- Unlike the results obtained by Ylisaukko-oja [25], thetem was found to work correctly up to 800 Hz by using a sin- experiments conducted in the home environment showedgle active sensor node with eight sensor inputs. However, the promising results as the system behaved the same way it didaddition of another active sensor node at this sampling during laboratory experiments. The presence of walls, floors,frequency resulted in an immediate loss of communication and possible sources of noise (home appliances) did notwith the receiver. increase the overall number of transmission errors. In fact,IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE JULY/AUGUST 2008 35
  8. 8. The rationale for in-home telerehabilitation is to extend rehabilitation services to the people in remote locations or with disabilities. 0.167% of transmission errors occurred during the reliability 1) The WBAN comprising four continuously streaming test and a mean value of 0.125% was obtained throughout the modules (100 Hz, eight channels per module) is the opti- house. A link can be made between the DFR and the number of mal configuration in terms of the number of active mod- communication losses. As expected, the further the WBAN is ules and communication errors. located from the receiver module, the greater the likelihood of 2) The WBAN when worn by an individual in a multilevel losing contact with the base station. Telerehabilitation sessions house during daily activities provides comparable per- usually take place in just one room where the camera, screen, formances and reliability as when in use under controlled and microphone are installed. Although not always the case, laboratory conditions. this study shows that the receiver unit could be located in 3) Data from the WBAN can be streamed over the Internet another room for wiring convenience. The embedded algo- without interfering with the performances of a videocon- rithms used for the experiments did not include any error man- ference link. agement functions. As explained previously, a missing data Future Applications and Challenges packet retransmission function could be embedded in the pro- Rehabilitation of patients with hip and knee replacements gram. Loss of communication also implies data loss. Every time usually involves the presence of a clinician for the assessment losses occur, the system must be restarted by consecutively of parameters such as joint range of motion. Remote assess- sending a ‘‘stop’’ and a ‘‘start’’ command. This process takes a ment of this parameter is possible using WBANs and acceler- few seconds and could be decreased to about 500 ms by auto- ometers. Wireless modules located at the patient’s ankle, mating the process. Although not desirable, occasionally miss- knee, and hip could serve as a goniometer providing angles for ing data packets is not critical during telerehabilitation sessions each segment using gravitational acceleration as a reference as long as the packets are correctly identified as missing. [10]. These measurements could help clinicians to better During the telerehabilitation session, sensors were wired to the assess their patients remotely. Combined signals from the sen- WBAN, and data were sent over the Internet to a remote site shar- sors, such as respiratory belts, pulse oximeter, and accelerom- ing the available bandwidth with a duplex audio and video signal eters, provide important information about a patient’s activity from a Tandberg system (Tandberg 550 MXP). The WBAN level during rehabilitation. The WBAN could remotely worked as expected, but some adjustments were made to transfer provide real-time data relating to patients’ exercise load and the data throughput from the sensors via the Internet to the clinical fatigue. This information could also be used to monitor site. Data reduction had to be done in order for the custom TCP/IP changes in patients’ health from one telerehabilitation session Labview application to work correctly and keep the connection to the next. Wireless weight-bearing is possible using instru- active. Data were filtered and downsampled three times before mented insoles wired to the WBAN. These sensors could be sending them to a remote computer, resulting in a bandwidth of used during telerehabilitation to evaluate gait and posture 209 kb/s (Figure 7). This bandwidth also includes polling the parameters and provide real-time feedback for the patient as router for statistics. Better results should be possible by allowing well as the clinician. Difficulties encountered during the home more lag in the communication and by establishing an error- telerehabilitation experiments provided important information managing algorithm for missing data packets. Overall, the telere- regarding design considerations for the next generation of habilitation session was not affected by the presence of the wireless platforms. The bracelets should be robust, comforta- WBAN system. Both clinical and home sites recorded great video ble, and easy for the patients to put on themselves [26]. The and sound performances throughout the session. Data from the docking station used for battery charging and remote program- sensors appeared on the clinician’s computer almost synchronized ming should be simple enough for the patients to clearly see with the video signals. From a telerehabilitation viewpoint, the that the modules are correctly docked. Overall, the study dem- battery life of the WBAN modules showed satisfactory results, as onstrated that remote monitoring from multiple sensor nodes they can be used extensively through the day and recharged at is technically feasible using a WBAN system and videocon- night using a docking station for charging batteries and remote ferencing system together. Future research will focus on data programming. Despite the manufacturer’s warnings about not reduction, choice of relevant sensors for remote assessment, using the transceivers within 1 m of each other, the system per- and software interfaces that will meet the emerging technical formed well in all dynamic tasks done by the subject. Special care guidelines for telehealth applications [27]. must be taken with the whip antenna that projects from the brace- let casing. These antennas will be replaced by smaller helix anten- Conclusions nas embedded directly in the bracelets. The use of a wireless body area network linked to embedded Experimental results obtained from laboratory and in-home and external sensors can increase the telepresence of rehabilita- testing of the proposed WBAN system can be synthesized as tion professionals by providing important information that is the following elements: otherwise difficult to obtain in that context. This article36 IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE JULY/AUGUST 2008
  9. 9. described the capability of a Zigbee-based WBAN and its poten- geriatric rehabilitation. He is currently funded as a Chercheurtial use in telerehabilitation applications. Experimental results Boursier Junior II by the Fonds de la recherche en sante du´show that a typical setup of four wireless sensor nodes with eight ´ Quebec.sensor inputs per node sampled at 100 Hz offers the most reli-able radio communication performance and reliability. Tests in Address for Correspondence: Patrick Boissy, Researcha real house showed the possibility of using the wearable system Centre on Aging, University Institute of Geriatrics of Sher-at home independently from the location of the receiver module brooke, Sherbrooke, Quebec, Canada. E-mail: patrick.and in conjunction with videoconferencing equipment. ReferencesWe wish to acknowledge the work done by Olivier Lessard- [1] L. Wakeford, P. P. Wittman, M. W. White, and M. R. Schmeler, ‘‘Telereha- `Fontaine and Simon Briere for the preliminary software bilitation position paper,’’ Am. J. Occup. Ther., vol. 59, no. 6, pp. 656–660,developments on the WBAN and videoconferencing soft- 2005. [2] J. M. Winters and J. M. Winters, ‘‘A telehomecare model for optimizingware. This project was funded by the Canadian Institutes for rehabilitation outcomes,’’ Telemed. J. E. Health., vol. 10, no. 2, pp. 200–212,Health Research, Institute of Musculoskeletal Health and 2004. [3] H. Hoenig, J. A. Sanford, T. Butterfield, and P. C. Griffiths, ‘‘Development ofArthritis (IMHA) under the Invention—Tools, Techniques a teletechnology protocol for in-home rehabilitation,’’ J. Rehabil. Res. Dev.,and Devices for Research and Medicine program (Grant No. vol. 43, no. 2, pp. 287–298, 2006.200307ITM-120560). [4] J. A. Sanford, P. C. Griffiths, P. Richardson, K. Hargraves, T. Butterfield, and H. Hoenig, ‘‘The effects of in-home rehabilitation on task self-efficacy in mobil- ity-impaired adults: A randomized clinical trial,’’ J. Am. Geriatr. Soc., vol. 54, Mathieu Hamel received his B.Eng. de- no. 11, pp. 1641–1648, 2006. ´ gree in 2003 from the Universite de Sher- [5] T. Sheridan, ‘‘Musings on telepresence and virtual presence,’’ Presence: Tele- oper. Virtual Environ., vol. 1, no. 1, pp. 120–126, 1992. brooke. In 2007, he completed his M.Sc. [6] W. Ijsselsteijn, H. de Ridder, J. Freeman, and S. Avons, ‘‘Presence: Concept, degree in electrical engineering in the field determinants and measurement,’’ presented at the International Society for Optical Engineering: Human Vision and Electronic Imaging, San Jose, CA, 2000. of signal processing at the same university. [7] M. Tousignant, P. Boissy, H. Moffet, and H. Corriveau, ‘‘In home telerehabili- He is now working at the Research Centre tation for older adults after discharge from an acute hospital or rehabilitation unit: on Aging, Sherbrooke, Canada, as a research A proof-of-concept study and costs estimation,’’ Disabil. Rehabil.: Assist. Tech- nol., vol. 1, no. 4, pp. 209–216, 2006. engineer. His professional interests include [8] H. Moffet, M. Tousignant, H. Corriveau, P. Boissy, S. Robitaille, F. Marquis,wireless devices applied to rehabilitation and biomedical signal and E. Anctil, ‘‘In home telerehabilitation after primary knee arthroplasty: A case report,’’ presented at the World Conf. Physical Therapy, Vancouver, Canada,processing. 2007. [9] J. M. Winters, Y. Wang, and J. M. Winters, ‘‘Wearable sensors and telereha- bilitation,’’ IEEE Eng. Med. Biol. Mag., vol. 22, no. 3, pp. 56–65, 2003. [10] H. Zheng, N. D. Black, and N. D. Harris, ‘‘Position-sensing technologies for ´ Rejean J. G. Fontaine received his B.Eng. movement analysis in stroke rehabilitation,’’ Med. Biol. Eng. Comput., vol. 43, ´ degree in 1991 from the Universite de Sher- no. 4, pp. 413–420, 2005. brooke. In 1999, he completed his Ph.D. [11] E. Jovanov, A. Milenkovic, C. Otto, and P. Groen, ‘‘A wireless body area network of intelligent motion sensors for computer assisted physical rehabilita- degree in electrical engineering at the same tion,’’ J. Neuroeng. Rehabil., vol. 2, no. 1, p. 6, 2005. university in the field of microelectronics [12] P. Lukowicz, T. Kirstein, and G. Troster, ‘‘Wearable systems for health care applications,’’ Methods Inf. Med., vol. 43, no. 3, pp. 232–238, 2004. applied to neurostimulation (cochlear im- [13] J. Sakari and N. Jarkko, Wireless Technologies for Data Acquisition Systems. plant). After a short stay in industry, he Dublin: Ireland: Trinity College Dublin, 2003. ´ returned to the Universite de Sherbrooke as [14] Zigbee-Alliance. (2006). ZigBee specifications [Online]. Available: http:// professor in 2001 and initiated works in electronics applied to [15] C. Otto, A. Milenkovic, A. Sanders, and E. Jovanov, ‘‘System architecture ofmedical imaging. He is now the chairman of a research group a wireless body area sensor network for ubiquitous health monitoring,’’ J. Mobileinvolved in the design of medical electronics dedicated to Multimedia, vol. 1, no. 4, pp. 307–326, 2006. [16] S. Victor, C. Bor-rong, L. Konrad, R. F. F. J. Thaddeus, and W. Matt, Sensorpositron emission tomography scanners and to biomedical Networks for Medical Care. San Diego, CA: ACM Press, 2005.signals. [17] Moteiv. (2006). Telos [Online]. Available: [18] K. Lorincz, D. J. Malan, T. R. F. Fulford-Jones, A. Nawoj, A. Clavel, V. Shnayder, G. Mainland, M. Welsh, and S. Moulton, ‘‘Sensor networks for Patrick Boissy received his B.Sc. degree emergency response: Challenges and opportunities,’’ IEEE Pervasive Comput. in kinesiology from the Universite de ´ Mag., vol. 3, no. 4, pp. 16–23, 2004. [19] MICAz. (2005). Crossbow technologies [Online]. Available: http://www. Sherbrooke in 1991. He graduated from ´ ´ the Universite de Montreal in 1999 with a [20] STMicroelectronics. (2004). LIS3L02AQ [Online]. Available: Ph.D. degree in biomedical sciences with com/stonline/products/literature/ds/9321.pdf [21] Murata. (2001). ENC-03M [Online]. Available: specialization in rehabilitation. After post- DATASHEET/giroscopio_CI/s42e3.pdf doctoral training at Boston University’s [22] ADInstruments. (2005). MLT1133 [Online]. Available: http://www.adinstruments. com/products/hardware/corporate/product/MLT1133/ Neuromuscular Research Centre, he joined [23] TinyOS. (2006). ‘‘Open-source operating system designed for wireless em- ´the faculty of the Universite de Sherbrooke in 2002 in bedded sensor networks,’’ [Online]. Available: http://www.zigbee.orgthe Kinesiology Department, where he is currently an associ- [24] NesC. (2003). nesC 1.1 language reference manual [Online]. Available: professor. He holds appointments as a researcher at the [25] A.-A. Ylisaukko-oja, A.-E. Vildjiounaite, and A.-J. Mantyjarvi, ‘‘Five-pointResearch Centre on Aging of the Health and Social Service acceleration sensing wireless body area network—Design and practical experien- ces,’’ presented at ISWC, 8th IEEE Int. Symp. Wearable Computers, Arlington,Centre, University Institute of Geriatrics of Sherbrooke, and VA, the Center of Excellence in Information Engineering of the [26] P. Bonato, ‘‘Advances in wearable technology and applications in physical ´Universite de Sherbrooke. His research interests include medicine and rehabilitation,’’ J. Neuroeng. Rehabil., vol. 2, no. 1, p. 2, 2005. [27] NIFTE. (2003). National initiative for telehealth framework of guidelines,technological and clinical evaluation of telehealth applica- [Online]. Available: and the study of the dose-response relationship in DocumentManager_op=downloadFileVV_File_id=54IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE JULY/AUGUST 2008 37