Multi-modal signal acquisition using a synchronized wireless body sensor network in geriatric patients
-
Maik Pflugradt
,
Steffen Mann
Abstract
Wearable home-monitoring devices acquiring various biosignals such as the electrocardiogram, photoplethysmogram, electromyogram, respirational activity and movements have become popular in many fields of research, medical diagnostics and commercial applications. Especially ambulatory settings introduce still unsolved challenges to the development of sensor hardware and smart signal processing approaches. This work gives a detailed insight into a novel wireless body sensor network and addresses critical aspects such as signal quality, synchronicity among multiple devices as well as the system’s overall capabilities and limitations in cardiovascular monitoring. An early sign of typical cardiovascular diseases is often shown by disturbed autonomic regulations such as orthostatic intolerance. In that context, blood pressure measurements play an important role to observe abnormalities like hypo- or hypertensions. Non-invasive and unobtrusive blood pressure monitoring still poses a significant challenge, promoting alternative approaches including pulse wave velocity considerations. In the scope of this work, the presented hardware is applied to demonstrate the continuous extraction of multi modal parameters like pulse arrival time within a preliminary clinical study. A Schellong test to diagnose orthostatic hypotension which is typically based on blood pressure cuff measurements has been conducted, serving as an application that might significantly benefit from novel multi-modal measurement principles. It is further shown that the system’s synchronicity is as precise as 30 μs and that the integrated analog preprocessing circuits and additional accelerometer data provide significant advantages in ambulatory measurement environments.
Introduction
Mobile devices designed to unobtrusively monitor various physiological parameters and motion states have emerged in considerable amounts and varieties during the past years [1, 11, 24, 34, 41, 47, 48, 72]. Typical fields of application range from acceleration-based activity monitors in smartphones to more sophisticated measurement setups in clinical environments. With reference to a medical context, the electrocardiogram (ECG), the photoplethysmogram (PPG), the electromyogram (EMG) and the electroencephalogram (EEG) span a group of very frequently monitored signals, along with further information on respiratory activity, body movements, skin conductance or temperature.
Next to heart rate, breathing rate and body temperature, the arterial blood pressure belongs to one of the primary vital signs. Whereas the auscultatory measurement approach is still considered the non-invasive standard in blood pressure detection [55], oscillometric and tonometric devices are also commonly applied in clinical applications [36, 57] but are not suited for continuous measurements, especially in ambulatory settings. The volume-clamp technique as employed by the Portapres system [5] has been proposed for continuous blood pressure monitoring, but exhibits some drawbacks due to its occlusive nature, leading to periods of venous congestion in long term recordings.
Thus, the demand for continuous and unobtrusive blood pressure monitoring alternatives remains highly prevailing, arousing a lot of current development and research activities in that field. A popular approach to assess the cardiovascular state and estimate arterial blood pressure is based on pulse wave velocity (PWV) considerations [37]. More specifically, non-obtrusive alternatives to blood pressure estimation exploiting the pulse arrival time (PAT) have been reported by [12, 43, 56]. Although these methods are in general insufficiently accurate according to absolute systolic blood pressure (SBP) and diastolic blood pressure (DBP) estimation, the dependency of PWV and SBP was still found to be statistically exploitable. PWV is often expressed in terms of PAT, which is closely connected to the pulse transit time (PTT) and is typically measured by the pulse wave traveling time to reach the peripheral arteries. The most common way to determine PAT is based on ECG triggered PWV detection or multiple site measurements incorporating PPG, pressure or bio-impedance signals [3, 9, 12, 19, 37].
The use of miniaturized and distributed wireless sensors shows its benefits not only in ambulatory settings but also in clinical environments. However, multi-modal signal acquisition procedures as PAT determination require a very accurate synchronization among the engaged sensors. The field of wireless synchronization mechanisms has been thoroughly researched including different radio protocols and architectures [7, 8, 14, 58, 60].
It is well known, that biosignal measurements are partially obstructed by different kinds of environmental influences and motion artifacts [62, 64]. Unlike in laboratory conditions, the subjects are usually not at rest and, especially in periods of patient movements, data acquisition devices like ECG and PPG sensors are seriously hampered due to unstable electrode-skin contacts or shifting photosensor positions. Moreover, changing circumstances such as electrical interference, ambient light, temperature, postures or activity levels have to be adequately taken into account when developing sensor systems or processing raw signals. To tackle these problems, we present a novel hardware system designed to enhance signal processing and evaluation approaches on data acquired in difficult environments. Integrated three-axis accelerometers comprise a major feature of the proposed system, providing continuous motion information that can be used to improve signal quality estimation or artifact reduction methods.
The aim of this work is twofold. First, a novel body sensor network will be thoroughly presented, giving a detailed insight into the technical parameters and performance. The implementational details and corresponding novelties are compared to recently published works. Second, the system’s applicability in clinical and ambulatory environments will be discussed. For that purpose, a clinical study to diagnose orthostatic hypotension in the elderly has been prepared, encompassing a Schellong test, which is a well-appreciated method to examine orthostatic dysregulations [70]. In that context, the ECG and PPG devices of the depicted hardware system will be used to record accurate PAT measurements.
Orthostatic hypotension, which is regarded as the most common disorder in blood pressure regulation after hypertension [59], is characterized by decreasing blood pressure following a change in body position from supine or seated to an upright posture. When standing up, approximately 750 ml of thoracic blood is immediately translocated downward. Normally, the SBP is expected to remain at a constant level as the body will compensate for the changing posture by increasing the heart rate. If this regulation is impaired to a certain degree, symptoms ranging from light-headedness to more severe cardiac episodes, and finally up to neurocardiogenic syncopes and cardiogenic shocks, can be provoked [59]. Factors promoting orthostatic hypotension have been found to originate from hypovolemia, disorders of the autonomic nervous system, vitamin B2 deficiency, vasodilation and general vessel diseases [18, 66]. As diagnostics of orthostatic intolerance are generally based on cuff blood pressure measurements which come along with the formerly mentioned difficulties, we propose to additionally exploit the benefits of multi-modal data analysis by accounting for unobtrusive PAT measurements based on the proposed system. This application will serve as a suitable example to discuss the virtues of the presented hardware system. However, it is not the scope of the paper to further depict the underlying signal processing approaches.
The paper is organized as follows. The section “Hardware” concerns itself with a detailed presentation of the robust body sensor network system. The synchronization mechanism is carefully depicted in the section “Wireless sensor synchronization and the experimental”. The setup and corresponding data acquisition are provided by the section “Experiments and PAT extraction”. The results will be presented in Section “Results” and the sections “Discussion” and “Conclusion” will conclude with a thorough discussion and short outlook.
Methods
Propelled by the ongoing miniaturization and availability of electrical components, numerous BSN hardware architectures have been presented in the last decade. The literature already offers vast amounts of wearable sensors that are battery powered and applied to biosignal or activity monitoring. Typically, the development of the corresponding systems concentrates on optimizing a specific parameter such as power consumption, size and weight, costs, maximum data rates, intelligent preprocessing or aspects like usability and unobtrusivness. Thus, it is not easy to get a complete overview and exhaustive comparison of the current advances.
Nonetheless, the works of Pantelopoulos and Bourbakis [47], Alemdar and Ersoy [1], Liolios et al. [34], Hao and Foster [24], Chen et al. [11] and Patel et al. [48] provide a very detailed overview of wearable health-monitoring systems published until 2012. Movassaghi [41] has written a great survey on current BSN achievements but concentrates more on general characteristics such as transmission protocols, channel models, security and data routing instead of comparing existing works with regards to specific applications.
After going into the details of our novel system in the following part, we will provide a thorough comparison with works depicted in the aforementioned reviews. We will also research recently published contributions to devise an updated picture presenting the current state of the art.
Hardware
To meet the initially mentioned requirements of a biomedical sensor network, a completely new system architecture will be presented in this section. The whole development has been driven by two main goals: First, to build a robust and reliable system, where accurate acquisition of raw signals can be guaranteed, and second, to provide as much comfort and functionality as possible with regards to clinical, laboratory as well as personal application environments. Therefore, the project has been termed robust body sensor network (rBSN).
To satisfy the demands of reliability, the internal data acquisition process incorporates elaborate control structures which detect and protocol run-time exceptions like lost data packets or deviations from timing intervals in the synchronization mechanism. Thus, each measurement is assigned a warranty seal confirming completeness and absence of any irregularities.
The aspect of usability also plays a major role, as the system should be designed to permit uncomplicated utilization by non technicians in clinical or private settings. Therefore, easy applicability and user-friendly interaction pose a significant requirement to the devices. Moreover, we aimed to design the sensors as modular as possible allowing for rapid development of novel sensors by recycling a majority of common modules. This way, the system is well suited for scientific applications, being designed for uncomplicated extensions and easy changes of the system’s behavior.
Technically, there are two major time critical-considerations that have to be accounted for to ensure data accuracy in wireless sensors. On the one hand, the complete data path, comprising analog preprocessing, digitization, buffering in RAM and final depositing on a mass storage device has to be handled in a given sampling interval. On the other hand, the time-crucial mechanism of timer synchronization has to be handled by the processor at highest priority. This is the reason, why a dual controller architecture as depicted in the left hand side of Figure 1 is proposed, physically separating the two mentioned tasks and thereby easing firmware development significantly.
(Left) Block diagram depicting rBSN sensor hardware based on dual controller architecture. All modules are powered by low dropout regulators (TI TPS72733) sourced by a 1300 mAh lithium-ion battery. (Right) Assembled rBSN_HeartCore sensor and its active electrodes.
Master:
The main controller (master) operates the analog front end and stores the digitized data on a micro SD-card. To buffer latencies of several milliseconds as they typically occur in write processes of FAT filesystems, the microcrontroller (μC) should offer enough SRAM along with powerful interfaces, high CPU frequencies and low power consumption. Therefore, the Texas Instruments MSP430F5659 (Texas Instruments, Dallas, TX, USA) has been chosen, offering 66 kb RAM, multiple serial SPI, I2C and UART interfaces, powerful timer, direct memory access, hardware CRC, hardware AES encryption and analog digital converter (ADC) modules, specifying a power consumption of only 1 mA/MHz at 3.3 V. The master runs an open FAT filesystem provided by Chan [10], handles user inputs and status LEDs, a serial USB interface based on an FTDIFT232 (Future Technology Devices International Ltd, Glasgow, UK) chip and controls communications with the slave controller. The master’s firmware actually consists of two parts. A high priority interrupt service routine triggers the corresponding ADC devices and writes the captured samples to an internal ping-pong buffer structure. The timer interrupt itself, which is used to define the sampling interval, is triggered by the slave controller presented in the following paragraph. Inside the lower prioritized main loop, full data buffers are safely written to the FAT filesystem when CPU resources become available.
Slave:
As mentioned earlier, the second controller (slave) is intended to spend all its resources operating a wireless radio module to implement timer synchronization and data streaming tasks. Concerning the wireless solution, the bluetooth (BT) specification was chosen for the following reasons. Firstly, BT has become a standard in commercial devices and is found in almost all smartphones, tablets and laptop computers. Therefore, implementation of sensor network handling, data visualization and configuration can be easily implemented resorting to standard hardware systems. Second, BT offers comparably generous bandwidths up to 2.1 Mbits/s so that multiple data channels can be streamed at high sampling rates up to several kilohertz. Moreover, BT has been successfully used for timer synchronization tasks [7, 8] and is suited for battery driven devices, as low power consumption modes have been introduced in the new low energy specification [4]. To handle these tasks, a MSP430F5438A controller constitutes the slave which runs a StoneStreetOne Blutopia Bluetooth Stack (Qualcomm, San Diego, CA, USA). The details of the synchronization mechanism are presented in the section “Wireless sensor synchronization” .
rBSN_HeartCore 12-Channel-ECG sensor node:
Research on measuring the electrical activity of the heart has started more than a century ago, when Willem Einthoven introduced one of the first commercial ECG devices. Since the 1960s, Holter monitors acquiring the body surface potential variations have been accepted as a well established method for ambulatory long time measurements allowing to track slowly varying parameters or seldom occurring events. Nonetheless, as soon as medical measurements are conducted in everyday situations, new obstacles like motion artifacts and further environmental influences affecting the signal quality have to be properly dealt with. Therefore, a novel ECG sensor based on the presented hardware architecture has been developed, incorporating new ideas to analog preprocessing and movement detection.
The onboard data acquisition is based on the ADS1298 single chip solution by Texas Instruments. This analog ECG front end provides an eight channel 24-Bit Delta-Sigma converter handling the signal digitization as well as numerous features like Wilson-Central-Terminal and Right-Leg-Driver potential generation. Using the eight input channels, a complete 12 channel ECG can be recorded, consisting of the three leads according to Einthoven I, II and III, the six precordial chest leads V1–V6 as well as the augmented limb leads aVR, aVL and aVF.
Before being routed to the onboard ADS1298 device, the signal is preprocessed by an active bandpass-circuitry. This is realized by a high impedance input buffer in the first stage, which is followed by a second order highpass filter removing the DC level at 0.05 Hz in the second stage. To reduce peaks caused by high capacitive loads, an external “in-the-loop” compensation has been implemented to enhance noise performance. Next, an active lowpass filter in multiple feedback architecture suppresses frequency components above 250 Hz. The fourth stage provides a non-inverting amplifier which further intensifies the signal’s amplitude. Due to the body surface potential’s amplitude, a thorough noise analysis has to be considered for the whole design. The assembled resistors and operational amplifiers can be seen as the major contributors of the expected system noise, which has been determined to be
rBSN_DualPulseOximeter:
Photoplethysmography has become a well-respected measurement approach to acquire information on the pulse wave traveling through the arterial tree [2, 69]. It has been deemed an indispensable standard to extract arterial oxygen saturation and pulse rates which are continuously monitored during anesthesia and post operative care. The PPG is typically recorded at extremities like the finger and earlobe (transmission mode) or on any measurement sites where there are large arteries like the radial or common carotid arteries (reflectance mode).
As introductory mentioned, the PPG can also be used to determine pulse wave velocities. Triggering the underlying time measurements with the help of the ECG R-peak is a typical way for PAT calculations, but demands for high accuracy in synchronicity, which is discussed in the section “Wireless sensor synchronization”.
We have developed a novel pulseoximeter, which is suited for measurements in transmission as well as reflectance mode. The main sensor node is based on the previously introduced system architecture. The finger clip houses five LEDs (wavelengths: 100 nm, 101 nm, 102 nm, 103 nm and 104 nm), a photodiode, a transimpedance amplifier as well as the three axis acceleration sensor ADXL335. The major details have been published in [53] and the benefits of exploiting the local acceleration signals were firstly discussed in [52].
rBSN_SkinConductance:
Electrodermal acitivity is a measure of skin conductivity, which varies depending on the sweat-induced moisture of the skin. It is usually measured at the palmar side of the hand or the bottom of the foot, as these regions hold a high concentration of eccrine sweat glands, whose activity is controlled by the sympathetic nervous system. A rise in sympathetic activity leads to the filling of sweat ducts, thereby increasing the skin conductivity. Therefore, electrodermal activity can be used as a convenient measure for the assessment of a person’s arousal. In combination with other biosignals it is furthermore used in the diagnosis and research of affective disorders, e.g. depression.
We extended the rBSN by a novel sensor that assesses the skin conductance through the constant-voltage-injection method. Here, a highly accurate voltage of 0.5 V is applied to the skin via two silver/silver-chloride electrodes. In order to measure both the phasic and tonic part of the electrodermal activity with maximum resolution, the high- and low-frequency components of the injected current are separated by an active filter network and individually measured by a transimpedance amplifier. The three axis accelerometer ADXL335 is also included for physical activity measurements.
rBSN_respBelt:
The current rBSN is complemented by a respiration acquisition device which provides the possibility to measure respirational activity in two different ways. For one, a standard piezo belt can be attached to the device which is worn around the chest or stomach. Secondly, an inductive measurement belt has been developed, based on a ferrite core that travels through a coil depending on thorax elongation. Respiration plays an important role, especially in sleep medicine, and can also be estimated by other biosignals such as the ECG or PPG [29, 44, 54]. The synchronized respiration belt sensor serve as a valuable reference for developing such novel signal extraction methods.
Wireless sensor synchronization
The importance of wireless timer synchronization in biomedical sensor networks has been repeatedly addressed in the previous sections and will be analyzed in the following part.
In wireless personal area networks (WPAN), Zigbee [14, 67], ANT [60] and bluetooth [8, 58] belong to the most common choices of wireless technologies. Whereas ANT and Zigbee offer relatively low maximum data rates (20 kbps and 250 kbps, respectively), the BT specification allows up to 2.1 Mbps. Thus, resorting to BT solutions provides considerably more bandwith for on-line data streaming applications. Furthermore, BT radio modules are broadly available and most commercial end devices like smartphones, tablets or laptops are equipped with a corresponding BT interface. Specifying a maximum current of 39.2 mA at full throughput and <1 mA in sleep and sniff modes, the TI CC256xbluetooth module is likewise suited for battery driven applications.
Casas et al. report a worst case synchronization error of 17.4 μs using the BT park state mechanism [8]. In a park state, the slave devices remain in an energy-saving sleep mode and synchronize on incoming broadcast messages at predefined beacon intervals. As the park state is not supported by all BT devices, the presented synchronization has been realized based on the BT sniff mode which will be depicted next. Based on the sniff mode Casamassima et al. have presented an accuracy of 313 ms [7].
BT communication in general is based on slots assigned a fixed length of 625 μs, used by master and slave devices to alternately transmit messages. Figure 2A shows a possible communication setup involving one master and two slave devices. When the high temporal resolution of 625 μs is not required, the BT sniff mode feature can be activated to drastically reduce power consumption. In sniff mode, each slave is sent to an inactive sleep state for a configurable period of time denoted as sniff interval, whose length is supposed to be a multiple of 625 μs. Having passed the sniff interval, the corresponding slave device wakes up and expects a message from the master. This specific time instance is also known as anchor point. In the presented timer synchronization mechanism, the sniff interval is set to 250 ms meaning that each slave expects a message from the master every 250 ms. When the slave receives the message, 250 ms have passed on the master’s clock. This way each slave device can synchronize its timer to the master clock to counteract present drift errors of the own crystals. Since a device’s relative anchor point position is known in the BT protocoll, the corresponding offset error can also be corrected. The synchronized timers are then used to trigger the sensor’s ADC modules, ideally resulting in perfectly synchronized samples of a multi-modal sensor application. Since the drift error accumulating in a short time window is negligible (<1 ms in 10 s), the proposed synchronisation mechanism is also robust towards lost BT packets.
![Figure 2: Bluetooth synchronization mechanism. (A) Underlying time division duplex mode on which bluetooth communication is based. Master and slaves can alternately send and receive messages [4]. (B) Timer drift correction implemented on the sensor device. After each sniff interval, the slave expects a message from the master. When this message is received, 250 ms have passed and the slave’s timer capture compare register (CCR) value can be adjusted accordingly to compensate the time delay denoted by TicksDrift.](/document/doi/10.1515/bmt-2014-0178/asset/graphic/j_bmt-2014-0178_fig_002.jpg)
Bluetooth synchronization mechanism. (A) Underlying time division duplex mode on which bluetooth communication is based. Master and slaves can alternately send and receive messages [4]. (B) Timer drift correction implemented on the sensor device. After each sniff interval, the slave expects a message from the master. When this message is received, 250 ms have passed and the slave’s timer capture compare register (CCR) value can be adjusted accordingly to compensate the time delay denoted by TicksDrift.
Experiments and PAT extraction
In the conducted clinical study, ten volunteers (70 years±10 years) participated in a cardiovascular measurement experiment involving a Schellong test [70]. All subjects provided a written consent in accord with the Ethics Board at Technical University of Berlin, from whom approval has been granted.
In each 15 min measurement, the patients were asked to lie on a bed, stand up and lie on a bed again for periods of 5 min each. A standard 12 lead ECG and a transmission pulseoximeter using the hardware described in the section “Hardware” were applied to synchronously acquire ECG and PPG signals at 500 Hz. Moreover, systolic and diastolic blood pressure, as well as oxygen saturation and pulse rate, were recorded at predefined time instances using a cuff based HEM-907XL Omron Healtcare blood pressure monitor.
The recorded data sets were stored in the European Data Format Plus (EDF+). With reference to PAT extraction, all pre-processing and calculation steps have been implemented in a standalone program called the rBSN_Viewer which is completely written in JAVA [51]. This tool offers a powerful chart component to plot the rBSN signals or standard EDF/EDF+ files from external databases. Moreover, annotating according to the EDF+ format is fully supported and can be significantly accelerated by built-in processing components like peak detectors or signal quality estimators. It is also possible to display on-line streaming data being sent by BT from the attached sensors. As JAVA provides the great advantage of platform independence, supporting a large range of hardware systems, the tool is also suited for various mobile devices.
Results
Hardware performance and synchronization accuracy
Hardware
The presented sensors were thoroughly tested to determine average power consumption and corresponding performance parameters. All sensors were supplied by a 3000 mAh lithium-ion-battery typically used in smartphones. The average power drawn by the rBSN_HeartCore sensor (fs =500 Hz) is around 110 mW, allowing for more than 27 h operating time.
Whereas a filesystem significantly eases post-data processing, it also limits the maximum amount of data that can be written to the micro SD-card. Using the mentioned FAT filesystem provided by ElmChan [10] and optimized ping-pong buffer structures, a throughput of 23.77 kbps without occurring data loss of the written data has been achieved. All samples belonging to a specific channel were stored in a packet structure with a 8 byte packet header as shown in Figure 3. The CRC checksum and packet count fields are used to verify recorded datasets for a specific channel. The packet length can be varied according to available space in RAM. Five hundred and four Bytes have proven to be a proper tradeoff. This length provides space to buffer data for more than 1 s (fs =500 Hz) and results in a reasonable header to data ratio of a single packet (1.5–98.5). With regards to the data throughput, a single channel could be sampled at 11.7 kHz or 24 channels at 500 Hz, respectively. It must be mentioned though, that the buffers require 40 kb RAM to guarantee undisturbed datastreams during sporadically occurring SD-card latencies, which are up to 500 ms. Such a reliable data acquisition is an essential prerequisite for multi modal signal processing approaches.
rBSN packet structure. A 16-bit CRC checksum and a 16-bit packet counter are used to verify data correctness. The number of data bytes are specified by packet length and the 1 byte channel id allows up to 255 channels in the whole BSN.
Synchronization
In order to quantitatively examine the synchronization accuracy an elaborate experiment, as depicted in Figure 4, has been conducted. Two sensors were attached to a frequency generator and configured to continuously record a sinewave (f=4 Hz, phase=0, offset=0, amplitude=3 V) at 500 Hz. When the synchronization mechanism is turned off, the sinewaves showed a phase mismatch of approximately 200 ms after 60 min, which is unacceptable for almost all biosignal applications (see Figure 4, left plot). This drift error is found to be in accordance with the worst case of 360 ms assuming a crystal accuracy of ±50 ppm.
Synchronization accuracy. (Left) The oscillator drift error is demonstrated in an elaborated experiment: Two sensor devices are sampling a sinewave provided by a frequency generator, with deactivated synchronization mechanisms. As can be seen in the fourth plot, the phase delay has accumulated to 250 ms within approximately 1 h, which is unacceptable for multi-modal biosignal applications. (Right) When the proposed synchronization mechanism is turned on, a timer accuracy of 30 μs has been achieved. The histogram shows the calculated phase delays between blockwise extracted windows of both sine waves, providing a standard derivation of 29.38 μs.
When the synchronization mechanism is turned on, it is supposed to completely eliminate drift and offset error. Therefore, the recorded sinewaves are analyzed with respect to their phase shifts. First, zero crossings are detected and then blocks of one second, which contain four sinewave periods, are extracted. An unconstrained, non-linear estimation technique is then applied to estimate the sinewave parameters of the extracted blocks [28]. The difference of the corresponding phase delays serves as a quantitative indicator describing the synchronization accuracy. The right hand side of Figure 4 shows the phase delays for several long-term measurements. According to the standard deviation, a synchronization accuracy of 30 μs has been achieved, which is acceptable for almost all standard biosignal applications.
The antenna directivity characteristics of the radio module have been investigated in several RX/TX experiments where the radio transmission path was obstructed by a water filled 10 l bucket, and lost packets have been analyzed with respect to varying sensor rotations (see Table 1). These measurements have been conducted in a laboratory environment with active WLAN routers and other BT devices operating in the same 2.4 GHz ISM frequency band. A mean lost packet rate of 9% was determined, which did not affect the synchronization accuracy.
Packet lost experiment for different sensor positions.
| Angle x | Angle z | Sent packets | Received packets | Packet lost |
|---|---|---|---|---|
| 0 | 0 | 1206 | 1077 | 10.69% |
| 45 | 0 | 1204 | 1085 | 9.88% |
| 90 | 0 | 1206 | 1105 | 8.37% |
| 135 | 0 | 1204 | 1083 | 10.04% |
| 180 | 0 | 1255 | 1148 | 8.52% |
| 225 | 0 | 1206 | 1090 | 9.61% |
| 270 | +0 | 1222 | 1095 | 10.39% |
| 315 | 0 | 1205 | 1079 | 10.45% |
| 0 | 45 | 1249 | 1140 | 8.72% |
| 0 | 90 | 1202 | 1090 | 9.31% |
| 0 | 135 | 1204 | 1086 | 9.8% |
| 0 | 180 | 1203 | 1087 | 9.64% |
| 0 | 225 | 1202 | 1068 | 11.14% |
| 0 | 270 | 1205 | 1076 | 10.70% |
| 0 | 315 | 1208 | 1100 | 8.94% |
The distance between the two sensors was 2 m. The radio path was additionally obstructed by a 10 l water bucket positioned between the sensors.
Comparison to related works
Although the literature provides an immense amount of works describing hardware designed for wireless biosignal acquisition, only a few contributions really present body area networks with wireless intra-sensor connections. For the sake of completeness, and to point out further advantages of the presented system, an extensive research of existing systems has been conducted. The results are listed in Table 2, which consists of three sections: Group A contains the highest scored systems analyzed in [47] and group B lists further significant papers drawn from all surveys [1, 11, 41, 47, 48]. As these reviews only cover submissions until 2012, Group C supplies recent works between 2013 and 2015 to get an as complete reference as possible. In the scope of the conducted research, WBANs utilizing wired synchronization mechanisms as well as systems consisting of only a single sensor have been discarded from further review [16, 21, 25, 27, 31, 63, 73].
Literature overview on true wireless body area networks.
| BSN | Signals | Synchronization accuracy | Mass storage | Clinical measurements |
|---|---|---|---|---|
| (A) Paper winners surveys | ||||
| MyHeart Project [23, 35] | ECG, ACC | n.a. | No | Yes |
| Rienzo 2005 [15] | ECG, Resp, ACC | n.a. | No | No |
| Pandian 2008 [46] | ECG, PPG, GSR, Temp | – | No | Yes |
| Human++ [22] | ECG, EEG, EMG | n.a. | No | Yes |
| Jin 2009 [26] | ECG, ACC | n.a. | No | No |
| Leijdekkers 2008 [20, 32] | ECG, BP, ACC | n.a. | No | No |
| (B) Works addressing wireless sync. | ||||
| Espina 2006 [17] | ECG, PPG | 28.4 μs | No | Yes |
| Milenkovic 2006 [38] | ECG, ACC, Footbutton | 30.5 μs | No | No |
| Monton 2008 [40] | ECG, ABP | 262.3 μs | No | No |
| Shnayder 2005 [61] | ECG, PPG, EMG, ACC | n.a. | No | Yes |
| Volmer 2008 [67] | ECG, PPG, PCG | 4 μs | Yes | No |
| (C) Relevant papers from 2013 to 2015 | ||||
| Caldara 2014 [6] | ACC, Gyro, Microphone Hygro., Barometer, Temp. | n.a. | No | Yes |
| Comotti 2013 [13] | Gyro, Geomagnetic | n.a. | No | No |
| LeMoullec 2014 [30] | Bioimpedance | 1 ms | No | No |
| Mo 2013 [39] | ACC | 6 ms | No | No |
| Peltokangas 2014 [49] | Force sensor, PPG, ECG, Respiration, | n.a. | No | No |
| Perez 2013 [50] | ECG, PPG, Respiration, Temp., | n.a. | No | No |
| Varatharajah2013 [65] | HeartRate, nasal air flow, ACC, Temp., | n.a. | No | No |
| Wang 2014 [68] | SpO2, ECG, ACC | n.a. | No | No |
| Xu 2013 [71] | ACC, PPG, Pressure | n.a. | No | No |
Group (A) lists the highest scored systems presented in [47], Group (B) contains further significant works drawn from the presented surveys and Group (C) names relevant works from the last 2 years. WBANs with wired synchronization or single sensor systems have not been considered in Group (B) and Group (C).
Pantelopulous and Bourbakis [47] assigned each reviewed hardware system a maturity score, which is based on different evaluation parameters, and also incorporates the patient’s, physician’s and manufacturer’s point of view. The best systems convince with regards to wearability (partly textile integrated sensors), aesthetic issues and their general applicability in real settings, as they have been successfully used in clinical studies. However, none of these works address the issue of wireless synchronization. The sensor woven clothes are indeed very suited for unobtrusive and ubiquitous measurements, but they might lack the flexibility for quick adjustments as required in laboratory experiments.
Group B provides true WBAN systems that clearly address the issue of synchronization and also evaluate the accuracy of the respective synchronization mechanisms. It should be noted, that these systems are based on the IEEE 802.15.4 standard and achieve very satisfactory synchronization errors in the range of several microseconds. Volmer and Orglmeister [67] is the only work considered in Table 2, that provides local mass storage capabilities. Unfortunately, no details regarding data file system, maximum data rates or reliability are given.
Searching for relevant papers that emerged in the past 2 years yielded only two contributions that touched upon synchronization issues. Mo et al. [39] present an ZigBee-based approach to synchronize two activity monitors with an accuracy of 6 ms. LeMoullec et al. [30] presents a system prototype that uses the IEEE 1588-2008 standard to synchronize multiple monitoring devices but provides only theoretical results.
PAT extraction during the Schellong test
Figure 5 shows an extract of the recorded signals during the conducted Schellong maneuver. ECG lead V5 and the fingerclip’s infrared pulse curve can be seen in the upper two plots. Plot 3 exemplarily shows two acceleration channels, the blue line belonging to ECG electrode V5 and the green to the PPG finger clip. The increased activity is clearly visible at t=220 s, when the subject moves from supine to standing posture. Plot 4 presents the extracted heart and pulse rates which do increase by 20 bpm. The blood pressure before standing up has been measured at 100/77 mm Hg and remained nearly constant at 95/70 mm Hg.
![Figure 5: Measurement signals 1) precordial ECG lead V5 with detected R-peaks, 2) transmission finger PPG with detected peaks [74] and automatic signal quality detection according to [33] plot, 3) X direction of acceleration channels from V5 electrode and PPG finger clip, 4) calculated heart and pulse rates plot, set to -1 on degraded PPG quality, 5) interpolated PAT.](/document/doi/10.1515/bmt-2014-0178/asset/graphic/j_bmt-2014-0178_fig_005.jpg)
Measurement signals 1) precordial ECG lead V5 with detected R-peaks, 2) transmission finger PPG with detected peaks [74] and automatic signal quality detection according to [33] plot, 3) X direction of acceleration channels from V5 electrode and PPG finger clip, 4) calculated heart and pulse rates plot, set to -1 on degraded PPG quality, 5) interpolated PAT.
The PAT is displayed in plot 5 and has been resampled at f=1 Hz using a linear interpolation method. In this study, two simple peak detectors have been applied to find the R-Peaks of ECG V5 [45] and onsets of the PPG [74] providing the necessary points for PAT calculation, which has decreased by approximately 50 ms during the change of posture in the depicted extract. Table 3 lists the relative changes of the extracted cardiovascular parameters obtained after changing from supine to standing posture from all subjects of the clinical study.
Changes of cardiovascular parameters (SBP, DBP, HR and PAT) due to the transition from supine to standing posture as has been conducted in the Schellong test.
| Patient | ΔSBP | ΔDBP | ΔHR | ΔPAT | ΔSBPe |
|---|---|---|---|---|---|
| F 83 years | +28 mm Hg | +9 mm Hg | +11 bpm | -7.8 ms | +11 mm Hg |
| F 84 years | +27 mm Hg | +14 mm Hg | +6 bpm | -33.2 ms | +7 mm Hg |
| F 65 years | +8 mm Hg | +0.5 mm Hg | -4 bpm | -17.1 ms | +2 mm Hg |
| F 79 years | -8.5 mm Hg | +7 mm Hg | +14 bpm | -61.4 ms | +1 mm Hg |
| F 80 years | -7 mm Hg | -16.5 mm Hg | +6 bpm | -6 ms | -1 mm Hg |
| F 85 years | +5 mm Hg | -1 mm Hg | -1 bpm | +15.0 ms | -3 mm Hg |
| F 84 years | +5 mm Hg | +2 mm Hg | +10 bpm | -61.8 ms | +1 mm Hg |
| M 83 years | +11 mm Hg | +6 mm Hg | +6 bpm | +9.1 ms | +19 mm Hg |
| F 87 years | -8 mm Hg | -1 mm Hg | +12 bpm | -30.2 ms | +5 mm Hg |
| F 84 years | +2 mm Hg | +6 mm Hg | +12 bpm | -20.0 ms | +2 mm Hg |
Using the derived PAT values and reference blood pressure measurements, Chen’s approach for continuous SBP estimation [12] has been applied to calculate the estimated SBP (SBPe) values which are listed in Table 3.
Moreover, an automatic signal quality estimator according to [33] has been implemented to detect motion artifacts. When a degraded signal quality is determined (indicated by a shaded background in the PPG plot), the current beat is excluded from pulse rate calculation as can be seen in the fourth plot, where the corresponding heart rates are set to -1. During longer periods of decreased signal quality, the PAT values will be labeled as invalid and can be excluded from further evaluation steps.
Discussion
In the scope of the presented field study, the system’s applicability has been successfully demonstrated in a clinical environment, thereby opening new possibilities for unobtrusive measurements. Especially diagnostics which depend on occlusive cuff blood pressure measurements might significantly benefit from the illustrated multi-modal methods. When a decrease in SBP and DBP can be reliably estimated based on PAT measurements as proposed by [12, 43, 56], the diagnostic procedure of detecting OH will be considerably simplified. Moreover, measurements conducted without an occlusive cuff element should have a lesser impact on the patient’s physiological state.
Different postures and movement activities were also shown to influence PAT measurements [42] and therefore have to be properly taken into account when evaluating the experiment. Particularly in ambulatory settings, the accelerometers integrated into each sensor provide a considerable advantage for post processing approaches, when signal artifacts can be related to present movements, as is indicated in Figure 5. The influence of different hand movements on automatic signal quality estimators applied on the PPG channel, has been investigated earlier in [52], giving an example of properly exploiting the additional movement information.
The presented signal evaluation procedure also contains an automatic signal quality detection step [33] in order to mark unreliable HR and PAT calculations. With regards to diagnose orthostatic hypotension, the PAT might serve as a valuable indicator for pathological classification. In subject 5, a mild form of instantaneous orthostatic hypotension has been diagnosed explaining the drop of SBP and the coincidental increase in heart rate, which are accompanied by a major decrease of PAT.
Conclusion
In this work, a novel wireless cardiovascular monitoring system has been presented, which is suitable for clinical as well as ambulatory environments and provides the flexibility needed for adjustments in elaborated scientific experiments. Although a lot of work has been done already, it is still hard to find a wireless body sensor network equipped with non wired synchronization mechanisms, reliable mass storage capabilities, proven applicability and long operating times on battery which is designed for unobtrusive biosignal acquisition. This work contributes a modular design to fill this gap in literature. It could be shown that using the bluetooth radio standard, multi-modal measurements with an underlying sensor synchronization accuracy of 30 μs are feasible for robust long-term recordings. Resorting to BT solution also offers the ability to connect the presented devices to commercial systems like smartphones, tablets and computers for immediate visualization, interaction and evaluation tasks. Moreover, the achieved throughput rates of 23.77 kbps on a FAT16 filesystem have proven sufficient to enable high resolution data acquisition at high sampling rates, which is required by various medical applications.
One of the major advantages of the proposed system is provided by the underlying modular architecture which enables easy adjustments on each device and allows for rapid development of new sensors that can be integrated to the rBSN network. Moreover, most parts of the firmware can be recycled, reducing liability to software flaws and easing version control as well as future updates.
To our best knowledge, no bluetooth synchronized WBAN system providing a complete 12 channel ECG, active electrodes equipped with acceleration sensors, reflexive as well as transmissive multi-wavelength PPG sensors, different respiration belts, electrodermal activity sensors and actimeters has been presented so far.
The system’s applicability has been successfully demonstrated in a clinical field study where the hardware has also proven its usability when operated by non-technicians. Equipping the measurement sites with well located acceleration sensors provides valuable information about the subject’s activity and contributes continuous data that can be used to assess the reliability of automatic signal processing approaches and novel evaluation methods.
In the next step, the automatic PAT extraction approach presented in this work will be further investigated to test the statistical significance for orthostatic hypotension diagnosis. Abstaining from occlusive blood pressure monitorings will not only enhance patient comfort and measurement accuracy, but also decrease medical attendance efforts and healthcare expenses. By exploiting the additional acceleration signals of the active electrodes, posture changes and movements can also be analyzed and properly taken into account for future PAT evaluation.
References
[1] Alemdar H, Ersoy C. Wireless sensor networks for healthcare: a survey. Comput Networks 2010; 54: 2688–2710.10.1016/j.comnet.2010.05.003Search in Google Scholar
[2] Allen J. Photoplethysmography and its application in clinical physiological measurement. Physiol Meas 2007; 28: R1.10.1088/0967-3334/28/3/R01Search in Google Scholar PubMed
[3] Bang S, Lee C, Park J, Cho M-C, Yoon Y-G, Cho SH. A pulse transit time measurement method based on electrocardiography and bioimpedance. In: Biomedical circuits and systems conference, BioCAS; Beijing, China: IEEE; 2009: 153–156.10.1109/BIOCAS.2009.5372060Search in Google Scholar
[4] Bluetooth Special Interest Group. Bluetooth core specification, core version 4.1. (dec. 2013). [online]. Available at: https://www.bluetooth.org/en-us/specification/adopted-specifications.Search in Google Scholar
[5] Bogert LWJ, van Lieshout JJ. Non-invasive pulsatile arterial pressure and stroke volume changes from the human finger. Exp Physiol 2005; 90: 437–446.10.1113/expphysiol.2005.030262Search in Google Scholar PubMed
[6] Caldara M, Comotti D, Galizzi M, et al. A novel body sensor network for parkinson’s disease patients rehabilitation assessment. In: 11th international conference on wearable and implantable body sensor networks (BSN); Zurich, Switzerland: 2014: 81–86.10.1109/BSN.2014.28Search in Google Scholar
[7] Casamassima F, Farella E, Benini L. Synchronization methods for bluetooth based wbans. In: IEEE international conference on body sensor networks (BSN); Cambridge, MA, USA: 2013: 1–6.10.1109/BSN.2013.6575489Search in Google Scholar
[8] Casas R, Gracia HJ, Marco A, Falco JL. Synchronization in wireless sensor networks using bluetooth. In: Third international workshop on intelligent solutions in embedded systems; Hamburg, Germany: IEEE; 2005: 79–88.Search in Google Scholar
[9] Cattivelli FS, Garudadri H. Noninvasive cuffless estimation of blood pressure from pulse arrival time and heart rate with adaptive calibration. In: Sixth international workshop on wearable and implantable body sensor networks BSN; Berkeley, California: IEEE; 2009: 114–119.10.1109/BSN.2009.35Search in Google Scholar
[10] ChaN. Fatfs – generic fat file system module (may. 2014) [online available] http://elm-chan.org/fsw/ff/00index_e.html.Search in Google Scholar
[11] Chen M, Gonzalez S, Vasilakos A, Cao H, Leung VC. Body area networks: a survey. Mobile Netw Appl 2011; 16: 171–193.10.1007/s11036-010-0260-8Search in Google Scholar
[12] Chen MW, Kobayashi T, Ichikawa S, Takeuchi Y, Togawa T. Continuous estimation of systolic blood pressure using the pulse arrival time and intermittent calibration. Med Biol Eng Comput 2000; 38: 569–574.10.1007/BF02345755Search in Google Scholar PubMed
[13] Comotti D, Ermidoro M, Galizzi M, Vitali A. Development of a wireless low-power multi-sensor network for motion tracking applications. In: IEEE international conference on body sensor networks (BSN); Cambridge, MA, USA: 2013: 1–6.10.1109/BSN.2013.6575466Search in Google Scholar
[14] Cox D, Jovanov E, Milenkovic A. Time synchronization for zigbee networks. In: Proceedings of the thirty-seventh southeastern symposium on system theory SSST’05; Tuskegee, AL, USA: IEEE; 2005: 135–138.Search in Google Scholar
[15] Di Rienzo M, Rizzo F, Parati G, Brambilla G, Ferratini M, Castiglioni P. Magic system: a new textile-based wearable device for biological signal monitoring. applicability in daily life and clinical setting. In: 27th annual international conference of the engineering in medicine and biology society IEEE-EMBS; Shanghai, China: 2005: 7167–7169.10.1109/IEMBS.2005.1616161Search in Google Scholar PubMed
[16] Dias R, J. Machado da Silva. A flexible wearable sensor network for bio-signals and human activity monitoring. In: 11th international conference on wearable and implantable body sensor networks workshops (BSN Workshops); Zurich, Switzerland: 2014: 17–22.10.1109/BSN.Workshops.2014.20Search in Google Scholar
[17] Espina J, Falck T, Muehlsteff J, Aubert X. Wireless body sensor network for continuous cuff-less blood pressure monitoring. In: 3rd IEEE/EMBS international summer school on medical devices and biosensors; Cambridge, MA, USA: 2006: 11–15.10.1109/ISSMDBS.2006.360085Search in Google Scholar
[18] Fan X-H, Wang Y, Sun K, et al. Disorders of orthostatic blood pressure response are associated with cardiovascular disease and target organ damage in hypertensive patients. Am J Hypertens 2010; 23: 829–837.10.1038/ajh.2010.76Search in Google Scholar PubMed
[19] Foo JYA, Wilson SJ, Dakin C, Williams G, Harris M-A, Cooper D. Variability in time delay between two models of pulse oximeters for deriving the photoplethysmographic signals. Physiol Meas 2005; 26: 531–44.10.1088/0967-3334/26/4/017Search in Google Scholar PubMed
[20] Gay V, Leijdekkers P. A health monitoring system using smart phones and wearable sensors’. Int J ARM 2007; 8: 29–35.Search in Google Scholar
[21] Ghoshdastider U, Viga R, Kraft M. Non-invasive synchronized spatially high-resolution wireless body area network. In: IEEE ninth international conference on intelligent sensors, sensor networks and information processing (ISSNIP); Singapore: 2014: 1–6.10.1109/ISSNIP.2014.6827593Search in Google Scholar
[22] Gyselinckx B, Vullers R, Van Hoof C, et al. Human++: Emerging technology for body area networks. In: IFIP international conference on very large scale integration; Nice, France: 2006: 175–180.10.1109/VLSISOC.2006.313229Search in Google Scholar
[23] Habetha J. The myheart project-fighting cardiovascular diseases by prevention and early diagnosis. In: Conf Proc IEEE Eng Med Biol Soc; Citeseer; 2006: 6746–6749.10.1109/IEMBS.2006.260937Search in Google Scholar PubMed
[24] Hao Y, Foster R. Wireless body sensor networks for health-monitoring applications. Physiol Meas 2008; 29: R27.10.1088/0967-3334/29/11/R01Search in Google Scholar PubMed
[25] Javaid N, Faisal S, Khan ZA, Nayab D, Zahid M. Measuring fatigue of soldiers in wireless body area sensor networks. In: Eighth international conference on broadband and wireless computing, communication and applications (BWCCA); Compiegne, France: 2013: 227–231.10.1109/BWCCA.2013.43Search in Google Scholar
[26] Jin Z, Oresko J, Huang S, Cheng AC. Hearttogo: a personalized medicine technology for cardiovascular disease prevention and detection. In: IEEE/NIH life science systems and applications workshop LiSSA; Bethesda, MD, USA: IEEE; 2009: 80–83.Search in Google Scholar
[27] Kantoch E, Augustyniak P, Markiewicz M, Prusak D. Monitoring activities of daily living based on wearable wireless body sensor network. In: 36th annual international conference of the IEEE engineering in medicine and biology society (EMBC); Chicago, IL, USA: 2014: 586–589.10.1109/EMBC.2014.6943659Search in Google Scholar PubMed
[28] Lagarias JC, Reeds JA, Wright MH, Wright PE. Convergence properties of the nelder–mead simplex method in low dimensions. SIAM J Optimiz 1998; 9: 112–147.10.1137/S1052623496303470Search in Google Scholar
[29] Langley P, Bowers EJ, Murray A. Principal component analysis as a tool for analyzing beat-to-beat changes in ECG features: application to ECG-derived respiration. IEEE Trans Biomed Eng 2010; 57: 821–829.10.1109/TBME.2009.2018297Search in Google Scholar PubMed
[30] Le Moullec Y, Lecat Y, Annus P, et al. A modular 6lowpan-based wireless sensor body area network for health-monitoring applications. In: Annual summit and conference Asia-Pacific signal and information processing association (APSIPA); Siem Reap, Cambodia: 2014: 1–4.10.1109/APSIPA.2014.7041570Search in Google Scholar
[31] Lee S-Y, Hong J-H, Hsieh C-H, Liang M-C, Chien S-YC, Lin K-H. Low-power wireless ECG acquisition and classification system for body sensor networks. IEEE J Biomed Health Inform 2015; 19: 236–246.10.1109/JBHI.2014.2310354Search in Google Scholar PubMed
[32] Leijdekkers P, Gay V. A self-test to detect a heart attack using a mobile phone and wearable sensors. In: 21st IEEE international symposium on computer-based medical systems CBMS’ 08; Jyväskylä, Finland: IEEE; 2008: 93–98.10.1109/CBMS.2008.59Search in Google Scholar
[33] Li Q, Clifford GD. Dynamic time warping and machine learning for signal quality assessment of pulsatile signals. Physiol Meas 2012; 33: 1491–501.10.1088/0967-3334/33/9/1491Search in Google Scholar PubMed
[34] Liolios C, Doukas C, Fourlas G, Maglogiannis I. An overview of body sensor networks in enabling pervasive healthcare and assistive environments. In: Proceedings of the 3rd international conference on PErvasive technologies related to assistive environments; Samos, Greece: ACM; 2010: 43.10.1145/1839294.1839346Search in Google Scholar
[35] Luprano J, Sola J, Dasen S, Koller J-M, Chetelat O. Combination of body sensor networks and on-body signal processing algorithms: the practical case of myheart project. In: international workshop on wearable and implantable body sensor networks BSN; Cambridge, MA, USA: 2006: 4–79.Search in Google Scholar
[36] Matthys K. Pascal Verdonck. Development and modelling of arterial applanation tonometry: a review. Technol Health Care 2002; 10: 65–76.10.3233/THC-2002-10107Search in Google Scholar
[37] McCombie DB, Reisner AT, Asada HH. Adaptive blood pressure estimation from wearable ppg sensors using peripheral artery pulse wave velocity measurements and multi-channel blind identification of local arterial dynamics. In: 28th annual international conference of the IEEE engineering in medicine and biology society EMBS ’06; New York, NY, USA: 2006: 3521–3524.10.1109/IEMBS.2006.260590Search in Google Scholar PubMed
[38] Milenković A, Otto C, Jovanov E. Wireless sensor networks for personal health monitoring: issues and an implementation. Comput Commun 2006; 29: 2521–2533.10.1016/j.comcom.2006.02.011Search in Google Scholar
[39] Mo L, Liu S, Gao RX, Freedson PS. Energy-efficient and data synchronized body sensor network for physical activity measurement. In: IEEE international instrumentation and measurement technology conference (I2MTC); Minneapolis, MN, USA: 2013: 1120–1124.10.1109/I2MTC.2013.6555588Search in Google Scholar
[40] Monton E, Hernandez JF, Blasco JM, et al. Body area network for wireless patient monitoring. IEEE Trans Commun 2008; 2: 215–222.10.1049/iet-com:20070046Search in Google Scholar
[41] Movassaghi S, Abolhasan M, Lipman J, Smith D, Jamalipour A. Wireless body area networks: a survey. IEEE Commun Surveys Tutorials 2014; 16: 1658–1686.10.1109/SURV.2013.121313.00064Search in Google Scholar
[42] Muehlsteff J, Aubert XA, Morren G. Continuous cuff-less blood pressure monitoring based on the pulse arrival time approach: the impact of posture. In: 30th Annual International conference of the IEEE engineering in medicine and biology society EMBS; Vancouver, BC: 2008: 1691–1694.10.1109/IEMBS.2008.4649501Search in Google Scholar PubMed
[43] Muehlsteff J, Aubert XL, Schuett M. Cuffless estimation of systolic blood pressure for short effort bicycle tests: the prominent role of the pre-ejection period. In: Conference proceedings:... Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Conference; volume 1, New York, NY: 2005: 5088–5092.10.1109/IEMBS.2006.260275Search in Google Scholar PubMed
[44] Nilsson L, Johansson A, Kalman S. Respiration can be monitored by photoplethysmography with high sensitivity and specificity regardless of anaesthesia and ventilatory mode. Acta Anaesth Scand 2005; 49: 1157–1162.10.1111/j.1399-6576.2005.00721.xSearch in Google Scholar PubMed
[45] Pan J, Tompkins WJ. A real-time qrs detection algorithm. IEEE Trans Biomed Eng 1985; 230–236.10.1109/TBME.1985.325532Search in Google Scholar PubMed
[46] Pandian PS, Mohanavelu K, Safeer KP, et al. Smart vest: wearable multi-parameter remote physiological monitoring system. Med Eng Phys 2008; 30: 466–477.10.1016/j.medengphy.2007.05.014Search in Google Scholar PubMed
[47] Pantelopoulos A, Bourbakis NG. A survey on wearable sensor-based systems for health monitoring and prognosis. IEEE Trans Syst Man CYbern C 2010; 40: 1–12.10.1109/TSMCC.2009.2032660Search in Google Scholar
[48] Patel S, Park H, Bonato P, Chan L, Rodgers M. A review of wearable sensors and systems with application in rehabilitation. J Neuroeng Rehabil 2012; 9: 21.10.1186/1743-0003-9-21Search in Google Scholar PubMed PubMed Central
[49] Peltokangas M, Vehkaoja A, Verho J, Huotari M, Roning J, Lekkala J. Monitoring arterial pulse waves with synchronous body sensor network. IEEE J Biomed Health Inform 2014; 18: 1781–1787.10.1109/JBHI.2014.2328788Search in Google Scholar PubMed
[50] Perez JJ, Saldarriaga AJ, Bustamante J. A wireless body sensor network platform to measure vital signs in clinical monitoring. In: Pan American health care exchanges (PAHCE); Medellin, Colombia: 2013: 1–6.10.1109/PAHCE.2013.6568301Search in Google Scholar
[51] Pflugradt M, Faerber T. rbsn_viewer (edf editing and annotating software) (oct. 2014) [online available] http://www.emsp.tu-berlin.de/menue/forschung/rbsn_viewer_edf_editing_and_annotating_software/.Search in Google Scholar
[52] Pflugradt M, Orglmeister R. Improved signal quality indication for photoplethysmographic signals incorporating motion artifact detection. In: 36th annual international conference of the IEEE engineering in medicine and biology society EMBS; Chicago, IL, USA: IEEE; 2014: 3422–3425.10.1109/EMBC.2014.6943975Search in Google Scholar PubMed
[53] Pflugradt M, Fritzsch I, Mann S, Tigges T, Orglmeister R. A novel pulseoximeter for bluetooth synchronized measurements in a body sensor network. In: 6th European DSP education and research conference (EDERC); Milano, Italy: 2014: 13–19.10.1109/EDERC.2014.6924351Search in Google Scholar
[54] Pflugradt M, Mann S, Feller V, Lu Y, Orglmeister R. Online learning algorithms for principal component analysis applied on single-lead ECGs. Biomed Tech/Biomed Eng 2013; 58: 121–130.10.1515/bmt-2012-0026Search in Google Scholar PubMed
[55] Pickering TG, Hall JE, Appel LJ, et al. Recommendations for blood pressure measurement in humans and experimental animals part 1: blood pressure measurement in humans: a statement for professionals from the subcommittee of professional and public education of the American Heart Association council on high blood pressure research. Hypertension 2005; 45: 142–161.10.1161/01.HYP.0000150859.47929.8eSearch in Google Scholar PubMed
[56] Poon CCY, Zhang YT. Cuff-less and noninvasive measurements of arterial blood pressure by pulse transit time. In: 27th annual international conference of the engineering in medicine and biology society IEEE-EMBS; Shanghai, China: 2005: 5877–5880.10.1109/IEMBS.2005.1615827Search in Google Scholar PubMed
[57] Ramsey M III. Blood pressure monitoring: automated oscillometric devices. J Clin Monitor 1991; 7: 56–67.10.1007/BF01617900Search in Google Scholar PubMed
[58] Ringwald M, Romer K. Practical time synchronization for bluetooth scatternets. In: Fourth international conference on broadband communications, networks and systems BROADNETS; Raleigh, NC, USA: IEEE; 2007: 337–345.10.1109/BROADNETS.2007.4550453Search in Google Scholar
[59] Romero-Ortuno R, Cogan L, Foran T, Kenny RA, Fan CW. Continuous noninvasive orthostatic blood pressure measurements and their relationship with orthostatic intolerance, falls, and frailty in older people. J Am Geriatr Soc 2011; 59: 655–665.10.1111/j.1532-5415.2011.03352.xSearch in Google Scholar PubMed
[60] Sheriff N. Time synchronization in ant wireless low power sensor network Thesis, University Jönköping, Sweden. 2011.Search in Google Scholar
[61] Shnayder V, Chen B-R, Lorincz K, Jones TRFF, Welsh M. Sensor networks for medical care. In SenSys, volume 5, pp. 314–314, 2005.10.1145/1098918.1098979Search in Google Scholar
[62] Such O. Motion tolerance in wearable sensors – the challenge of motion artifact. In: 29th annual international conference of the IEEE engineering in medicine and biology society EMBS; Lyon, France: 2007: 1542–1545.10.1109/IEMBS.2007.4352597Search in Google Scholar PubMed
[63] Sun F, Zhao Z, Fang Z, Chen D, Chen X, Xuan Y. Design and implementation of an ultra low power health monitoring node for wireless body sensor network. In: Fourth international conference on digital manufacturing and automation (ICDMA); Qingdao, China: 2013: 417–422.10.1109/ICDMA.2013.98Search in Google Scholar
[64] Sweeney KT, Ward TE, McLoone SF. Artifact removal in physiological signals – practices and possibilities. IEEE Trans Inf Techn Biomed 2012; 16: 488–500.10.1109/TITB.2012.2188536Search in Google Scholar PubMed
[65] Varatharajah Y, Karunathilaka N, Rismi M, Kotinkaduwa S, Dias D. Body area sensor network for evaluating fitness exercise. In: 6th Joint IFIP wireless and mobile networking conference (WMNC); Dubai: 2013: 1–8.10.1109/WMNC.2013.6548957Search in Google Scholar
[66] Verwoert GC, Mattace-Raso FUS, Hofman A, et al. Orthostatic hypotension and risk of cardiovascular disease in elderly people: the rotterdam study. J Am Geriatr Soc 2008; 56: 1816–1820.10.1111/j.1532-5415.2008.01946.xSearch in Google Scholar PubMed
[67] Volmer A, Orglmeister R. Wireless body sensor network for low-power motion-tolerant syncronized vital sign measurment. In: 30th annual international conference of the IEEE engineering in medicine and biology society EMBS; Vancouver: 2008: 3422–3425.10.1109/IEMBS.2008.4649941Search in Google Scholar PubMed
[68] Wang Z, Zhao C, Qiu S. A system of human vital signs monitoring and activity recognition based on body sensor network. Sensor Rev 2014; 34: 42–50.10.1108/SR-12-2012-735Search in Google Scholar
[69] Webster JG. Design of pulse oximeters. CRC Press 1997.10.1887/0750304677Search in Google Scholar
[70] Winker R, Prager W, Haider A, Salameh B, Rüdiger HW. Schellong test in orthostatic dysregulation: a comparison with tilt-table testing. Wien Klin Wochenschr 2005; 117: 36–41.10.1007/s00508-004-0288-5Search in Google Scholar PubMed
[71] Xu G, Li JQ, Wang L, et al. Hardware design of a body sensor network system used for elder care. In: Ninth international conference on natural computation (ICNC); Shenyang, China: 2013: 1585–1589.10.1109/ICNC.2013.6818234Search in Google Scholar
[72] Yang G-Z, Yacoub M. Body sensor networks. Springer; 2006.10.1007/1-84628-484-8Search in Google Scholar
[73] Zhou C, Tu C, Tian J, et al. A low power miniaturized monitoring system of six human physiological parameters based on wearable body sensor network. Sensor Rev 2015; 35: 210–218.10.1108/SR-08-2014-687Search in Google Scholar
[74] Zong W, Heldt T, Moody GB, Mark RG. An open-source algorithm to detect onset of arterial blood pressure pulses. In: computers in cardiology; Thessaloniki Chalkidiki, Greece: IEEE; 2003: 259–262.10.1109/CIC.2003.1291140Search in Google Scholar
©2016 by De Gruyter
Articles in the same Issue
- Frontmatter
- Editorial
- Biosignal processing
- Review
- A review of beat-to-beat vectorcardiographic (VCG) parameters for analyzing repolarization variability in ECG signals
- Research articles
- Classification of persistent and long-standing persistent atrial fibrillation by means of surface electrocardiograms
- Entropy at the right atrium as a predictor of atrial fibrillation recurrence outcome after pulmonary vein ablation
- P wave detection and delineation in the ECG based on the phase free stationary wavelet transform and using intracardiac atrial electrograms as reference
- Multi-modal signal acquisition using a synchronized wireless body sensor network in geriatric patients
- A portable device for recording evoked potentials, optimized for pattern ERG
- Random forests in non-invasive sensorimotor rhythm brain-computer interfaces: a practical and convenient non-linear classifier
- Fractal and twin SVM-based handgrip recognition for healthy subjects and trans-radial amputees using myoelectric signal
- Nonlinear analysis of pupillary dynamics
- A multichannel bioimpedance monitor for full-body blood flow monitoring
- Recognition of amyotrophic lateral sclerosis disease using factorial hidden Markov model
- Short communication
- Quantifying the complexity of human colonic pressure signals using an entropy measure
Articles in the same Issue
- Frontmatter
- Editorial
- Biosignal processing
- Review
- A review of beat-to-beat vectorcardiographic (VCG) parameters for analyzing repolarization variability in ECG signals
- Research articles
- Classification of persistent and long-standing persistent atrial fibrillation by means of surface electrocardiograms
- Entropy at the right atrium as a predictor of atrial fibrillation recurrence outcome after pulmonary vein ablation
- P wave detection and delineation in the ECG based on the phase free stationary wavelet transform and using intracardiac atrial electrograms as reference
- Multi-modal signal acquisition using a synchronized wireless body sensor network in geriatric patients
- A portable device for recording evoked potentials, optimized for pattern ERG
- Random forests in non-invasive sensorimotor rhythm brain-computer interfaces: a practical and convenient non-linear classifier
- Fractal and twin SVM-based handgrip recognition for healthy subjects and trans-radial amputees using myoelectric signal
- Nonlinear analysis of pupillary dynamics
- A multichannel bioimpedance monitor for full-body blood flow monitoring
- Recognition of amyotrophic lateral sclerosis disease using factorial hidden Markov model
- Short communication
- Quantifying the complexity of human colonic pressure signals using an entropy measure
