Multimedia in Healthcare.pptx

Everything related to Multimedia in Healthcare is clearly explained and will help you solve all your problems on how important Multimedia is in Healthcare.

1. University of Information and Communication Technology Faculty of Engineering Department of Information and Communication Technology Engineering Multimedia in Healthcare Eng -   Fatima Asaad Hikmat Al- zaidi 2022/10/12

2. بسم الله الرحمن الرحيم هُوَ الَّذي جَعَلَ الشَّمسَ ضِياءً وَالقَمَرَ نورًا وَقَدَّرَهُ مَنازِلَ لِتَعلَموا عَدَدَ السِّنينَ وَالحِسابَ ما خَلَقَ اللَّـهُ ذلِكَ إِلّا بِالحَقِّ يُفَصِّلُ الآياتِ لِقَومٍ يَعلَمونَ صدق الله العظيم

3. Inscription. Man was created on earth, and did not live in isolation from the rest of humanity and at all stages of life there are people who deserve thanks and the first people to be guided are the fathers because of their credit for the heavens, their presence is a reason for survival and peasantin our parents and the hereafter, so I dedicate this humble research to them first and then to my husband and companion of the struggle in the course of life and his support for me in my studies. I dedicate my distinguished professor, Brof . Dr. Abbas Fadhil al- Jubouri , and all the teachers of the Department of Information Technology Engineering, and to all those who stood by me and supported me, even with the encouragement of only my brothers, friends and friends, to whom I testify that they are the blessings of the comrades in all things.

4. Thanks and appreciation. the first thanks is Allah Almighty, then my parents and my mother for all their efforts since I was born to this moment, you are everything I love you in God the most love. I am pleased to thank all those who have advised me, guided me, guided me or contributed to the preparation of this research by communicating with me the required references and sources at any stage. In particular, I thank my distinguished professor, Dr. Abbas Fadel Al- Jubouri , for teaching me and guiding me and for choosing the topic. I also thank all the professors of the Department of Information Technology Engineering, .

5. Contents 1. Abstract. 2. Introduction. 3. A remote marker less human gait tracking for e- healthcare based on content-aware wireless multimedia communications. 4. Marker less human gait tracking. 5. Where the denotes a threshold. 6. Wireless streaming of remote human gait tracking the proposed system model. 7. The joint optimization of content-aware wireless streaming .

6. 8. System experiments. 9 . Conclusions . 10.A multimedia telemonitoring network for healthcare. 11.TELMES project aims to develop a securized multimedia system devoted to medical. 12.Telemonitoring module. 13.Results and discussion. 14.Summary and conclusion. 15.Multimedia-based healthcare. 16.References .

7. Abstract As the world population ages and the care ratio (ratio of healthy young citizens to elderly citizens) is in decline, monitoring and care of individuals with continuous recording of medical information in electronic form remotely or during in-site medical visits is becoming more and more incorporated in daily life. Multimodal technologies are constantly being developed to assist people in their daily life, with a vast range of wearable sensors now available for monitoring health parameters (e.g. blood pressure, sweat, body temperature, heart rate etc.), Multimedia in Healthcare

8. lifestyle (e.g. monitoring utility use, levels of activity, sleep quantity and quality etc.), a person’s ability to carry out activities of daily living. At the same time, health professionals have integrated new technologies into their workflow, for example by using various types of medical imagery to facilitate and support their clinical practice and diagnosis, and also by examining data from sensors and home medical devices, which allow them to remotely care for their patients. Health records and databases are now enriched with digital multimodal data on the patients, for which new methods need to be developed for accurate and fast access and retrieval.

9. Introduction It is reported that certain diseases can directly affect human gait, stride, pace, and overall walking pattern. Human gait tracking, the process of identifying an individual by walking manner, can then be widely used in surveillance, biomedical assistance, physical training and therapy. For instance, it can readily be used to provide prognostic and diagnostic measures of pathological locomotion biorhythm such as Parkinson's disease, diabetic peripheral neuropathy, and Huntington's disease.

10. It can also be utilized for the clinical assessment of stroke rehabilitation, prosthetic alignment, and the success of orthopedic interventions such as anterior cruciate ligament reconstruction . Therefore , human motion tracking and gait analysis create an opportunity for early intervention and possibly the prevention of injury as a result of declining condition from disease or the natural aging process [1, 2].

11. However , most existing tracking systems usually rely on expensive equipment and lengthy processes to collect gait data in a dedicated biomechanical environment, limiting their accessibility to small clinics located in remote areas . On the other hand, the increasing prevalence of inexpensive hardware such as video phones . And complementary metal oxide semiconductor (CMOS) cameras has fostered the development of wireless multimedia technologies, such as wireless multimedia sensor networks (WMSNs).

12. WMSNs not only enhance existing sensor net- work applications such as tracking, surveillance, home automation, and environmental monitoring, but also facilitate new medical applications such as telemedicine and advanced health care delivery . Indeed , telemedicine is a fast-growing application of clinical medicine, where medical information is transferred via telephone, the Internet, or wireless networks for the purpose of consulting, and sometimes remote medical procedures or examinations.

13. Telemedicine can be integrated with third-generation (3G) broadband multimedia networks to provide ubiquitous e - healthcare services. Furthermore, remote monitoring is a new technology emerging to improve disease treatment and lower medical costs. The essence of remote monitoring is to enable assessment of an individual's medical status in real time regardless of his or her location, and to allow a doctor or a computer to view the information anywhere to aid diagnosis, observe how a treatment is working, or determine if a condition has become acute.

14. Therefore, an e-healthcare platform based on wireless multimedia technologies for remote marker less human motion tracking and gait analysis will significantly improve current clinical medicine practices, especially for small clinics located in remote areas. Patients in remote areas can also get online screening or a preliminary diagnosis before they are physically transported to a fully equipped medical center to save time and resources. By reducing the number of visits to clinics or care facilities, medical costs can be significantly reduced, and convenience and care quality are highly improved.

15. In this article we present a new marker less human gait tracking based on content-aware wireless streaming for remote tracking, as shown in Fig. 1 . The proposed system first extracts the human motion region accurately by jointly utilizing properties of interframe relations, as well as spectral and spatial inter-pixel-dependent con- texts of the video frames . Thus , it does not need the expensive equipment in the traditional pre- designed intrusive marker-based environment (Fig. 2).

16. Then , based on content aware analysis results, collected gait data are transmitted to the medical center for speedy or real-time medical prognosis and diagnosis through wireless net- works. A proposed quality-driven distortion-delay framework is adopted to guarantee user-perceived video quality at the receiver end. Specifically, video coding parameters at the application layer, and a modulation and coding scheme at the physical layer are jointly optimized through a minimum distortion problem to satisfy a given playback delay deadline.

17. In other words, the optimal combination of encoder parameters and adaptive modulation and coding (AMC) scheme are chosen to achieve the best fidelity of the collected gait data. Also , the extracted human gait region can be coded with finer parameters, and be transmitted with better protection than the insignificant background area under the given quality of service ( QoS ) requirement.

18. Experimental results using H.264/AVC have shown the validity and effectiveness of the proposed system . The remainder of this article is organized as follows. We present the procedures for marker less human motion tracking . We also discuss the proposed optimized delivery of wireless streaming based on the content-aware analysis results of the previous section . We then describe the system environment as well as the experimental results. Finally, we conclude the article.

19. A remote marker less human gait tracking for e-healthcare based on content-aware wireless multimedia communications Remote human motion tracking and gait analysis over wireless networks can be used for various healthcare systems for fast medical prognosis and diagnosis. However , most existing gait tracking systems rely on expensive equipment and take lengthy processes to collect gait data in a dedicated biomechanical environment, limiting their accessibility to small clinics located in remote areas.

20. In this work we propose a new accurate and cost-effective e- healthcare system for fast human gait tracking over wireless networks, where gait data can be collected by using advanced video content analysis techniques with low-cost cameras in a general clinic environment.

21. In this way the encoder behavior and the modulation and coding scheme are jointly optimized in a holistic way to achieve the best user-perceived video quality over wireless networks. Experimental results using H.264/AVC demonstrate the validity and efficacy of the proposed system. Furthermore, based on video content analysis, the extracted human motion region is coded, transmitted, and protected in video encoding with a higher priority against the insignificant background area to cope with limited communication bandwidth .

22. Marker less Human Gait Tracking Given a gait video sequence, the human gait region contains valuable information for gait analysis and medical examinations. In contrast, the background area usually provides much less useful knowledge . However , according to our study, for a video frame of the collected human gait data, the background area usually comprises more than 50 percent of the whole video frame area.

23. Transmitting the background area and the human gait region without differentiation is not only unnecessary but also wasteful of precious wireless network resources. Therefore , the identification of human gait region through the proposed content-aware analysis techniques is a meaningful task. The result provides a foundation for wireless transmission of remote human gait tracking by prioritizing the human gait region against the unimportant background area under the bandwidth-limited environment.

24. Further- more, the proposed marker-less human motion tracking system makes gait tracking much less dependent on dedicated gait tracking facilities.

25. Figure 3 shows the proposed procedures for marker less human gait tracking and the corresponding results, where background subtraction and contextual classification play major roles. Through these two steps, the temporal inter frame correlation of the video clip, the spectral and spatial inter-pixel dependent contexts of the video frames are cooperatively utilized, making it possible to accurately track the human gait region in a marker less environment.

26. Background Subtraction- The rationale of back ground subtraction is to detect the moving objects from the difference between the current frame and a reference frame, often called the "background image. Therefore , the background image is ideally a representation of the scene with no moving objects and is kept regularly updated so as to adapt to the varying luminance conditions and geometry settings [3].

27. Given a video sequence, the objective of background subtraction is to detect all foreground objects (i.e., the human gait region in this article). The naive description of the approach is to depict the human gait region as the difference between the current frame Fri and the background image Bgi :

28. where The denotes a threshold However, the background image is not totally static. Therefore, before this approach can actually work, certain factors need to be adapted to, which includes illumination changes, motion changes, as well as sometimes changes in back- ground geometry. Over time, different back- ground objects are likely to appear at the same pixel location. Sometimes the changes in the background object are not permanent and appear at a rate faster than that of the background update .

29. To model this scenario, a multivalued background model can be adopted to cope with multiple background objects . Therefore , algorithms such as the proposed Gaussian Mixture Model (GMM) can define an image model more properly as it provides a description of both fore- ground and background values [4]. The result of this step using GMM is illustrated in Fig. 3b. Contextual Classification- The objective of classification is to classify a video frame by the object categories it contains.

30. Supervised classification is a type of automatic multispectral image interpretation in which the user supervises feature classification by setting up prototypes (collections of sample points) for each feature class to be mapped. A supervised contextual classification that utilizes both spectral and spatial contextual information can better discriminate between pixels with similar spectral attributes but located in different regions. First, in many images, especially those remotely sensed images, object sizes are much greater than the pixel element size.

31. There- fore, the neighboring pixels are more likely to belong to the same class, forming a homogeneous region. Furthermore, some classes have a higher possibility of being placed adjacently than others, so the information available from the relative assignments of the classes of neighboring pixels is also very important. By using both spectral and spatial contextual information, the speckle error can effectively be reduced, and the classification performance can be improved significantly. Nonetheless, this type of classification also suffers from the problem of small training sample size, where the class conditional probability has to be estimated in the analysis of hyper spectral data. Therefore algorithms such as adaptive Bayesian contextual classification that utilizes both spectral and spatial inter-pixel-dependent contexts to estimate the statistics and classification can be adopted for accurate classification [5].

32. This model is essentially the combination of a Bayesian contextual classification and an adaptive classification procedure. In this classification model only inter-pixel class dependency context is considered, while the joint prior probabilities of the classes of each pixel and its spatial neighbors are modeled using Markov random fields (MRF) [6]. The result of this step is shown in Fig. 3c. Region Growing- At this stage, shown in Fig. 3d, density check is adopted to combine the results of the previous two steps to form the continuous human gait region. As long as a homogeneous region achieved from stage 2 contains more than a threshold percentage of human motion pixels obtained from stage 1, this region is regarded as the human gait region. Otherwise, it falls into the background area.

33. Morphological Operations and Geometric Corrections - Results from the previous stage contain undesired noises and holes. As shown in Fig. 3e, morphological operations use dilation and erosion to populate the holes in the human motion region and remove the small objects in the background areas. Then, geometric correction can be performed horizontally and vertically to further remove noises for the accurate achievement of human gait [7]. Finally, we achieve the desired results of the extracted human gait region as displayed in Fig. 3f.

34. Wireless Streaming of Remote Human Gait Tracking The Proposed System Model Based on the proposed marker less human motion tracking results, we also develop a unified quality- driven optimization system of wireless streaming for delay-bounded human gait transmission. Figure 4 illustrates the proposed system model, which consists of an optimization controller, the marker- less human gait tracking module, a video encoding module, as well as the modulation and coding module. To increase the overall video quality, we first adopt the proposed methods mentioned earlier to identify the human gait region in a marker- less environment. Due to the different contribution to gait analysis from the human motion region and the background area, the human gait region can be coded at finer quality at the video encoder of the application

35. layer, and the packets of the human gait region can use smaller constellation sizes and lower channel coding rates to guarantee the required packet error rate at the physical layer. By redistributing the limited resources needed for encoding and transmission according to the video content, the overall quality of real-time human gait video delivery over wire- less networks will be significantly improved. Furthermore, the controller is the core of the proposed system, which can jointly optimize the key system parameters of the video codec at the application layer, and the modulation and coding schemes at the physical layer. Therefore, through these parameters, the controller can control the behaviors of the video encoder and the modulation and coding module.

36. More important, adjusting with coordination the system parameters of the video encoder residing in the application layer and the modulation and coding scheme residing in the physical layer can greatly enhance overall network performance. As shown in Fig. 5, the sys- tem performance in terms of video distortion is jointly decided by the encoder behavior (i.e., quantization step size, or QP) and the packet loss rate. Additionally, packet loss rate is determined by bit error rate (BER), which is then collectively affected by the channel quality and AMC scheme. Therefore, all related system parameters can be holistically optimized toward achieving the best possible video quality under a given delay constraint.

37. For example, when a wireless channel is experiencing bad quality, the time-varying channel information can be used to dynamically adapt the AMC scheme to minimize the packet loss rate, enhancing the received video quality over wireless networks. Thus, the proposed cross-layer joint optimization is able to choose the optimal set of parameter values to achieve the best received video performance, providing a natural solution to improve overall system performance for wire- less streaming of remote human gait tracking.

38. The Joint Optimization of Content-Aware Wireless Streaming At the video encoder, for hybrid motion-compensated video coding and transmission over loss channels, each video frame is generally represented in block-shaped units of the associated luminance and chrominance samples (16 x 16 pixel region) called macro blocks (MBs). In the H.264 codec MBs can be either intra-coded or inter-coded from samples of previous frames [8]. Intra-coding is performed in the spatial domain, by referring to neighboring samples of previously coded blocks to the left and/or above the block to be predicted.

39. Inter-coding is performed with temporal prediction from samples of previous frames. Many coding options exist for a single MB, and each of them provides different rate- distortion characteristics. In this work only pre- defined MB encoding modes are considered, since we want to apply error-resilient source coding by selecting the encoding mode of each particular MB. This is crucial to allow the encoder to trade off bit rate with error resiliency at the MB level. For real-time source coding, the estimated distortion caused by quantization, packet loss, and error concealment at the encoder can be calculated by using the Recursive Optimal Per-Pixel Estimate (ROPE) method, which provides an accurate video-quality-based optimization metric to the cross-layer optimization controller [9].

40. At the physical layer, the BER pem is decided by the dynamically chosen mode of AMC, which has been advocated to enhance the throughput of future wireless communication systems at the physical layer [10]. With AMC, the combination of different constellations of modulation and different rates of error control codes are chosen based on the time-varying channel quality. For example, in good channel conditions, an AMC scheme with larger constellation sizes and high channel coding rate can guarantee the required packet error rate, which means that AMC can effectively decrease the transmission delay while satisfying the packet loss rate construing. Each AMC mode consists of a pair of modulation scheme a and FEC code c as in Third Generation Partnership Project (3GPP), HIPERLAN/2, IEEE 802.lla, and IEEE 802.16 standards. Furthermore, we adopt the following approximated BER expression:

41. where m is the AMC mode index and is the received SNR. Coefficients am and bum are obtained by fitting Eq. 2 to the exact BER as shown in Fig. 5. Therefore, the expected mean squared error (MSE) between the received pixels and original pixels of the video frames can be adopted as the

42. distortion metric [9]. Thus, the expected distortion Ed(ρ) accurately calculated by ROPE under instantaneous network conditions, which is represented by packet loss rate ρ, becomes the objective function in the proposed optimization framework. Packet loss rate ρ can be further calculated from BER pme as long as the packet size is known. Meanwhile, the transmission delay ttrans which is constrained by the given frame delay bound Tf , can be represented by bandwidth Be and data bit rate Rm. Finally, the problem can be formulated as a minimum distortion problem constrained by a given frame delay bound.

43. By eliminating from the potential solution set the parameters that make the transmission delay exceed the delay constraint, the constrained problem can be relaxed to an unconstrained optimization problem. Furthermore, most decoder concealment strategies introduce dependencies among slices. For example, if the concealment algorithm uses the motion vector of the previous MB to conceal the lost MB, it would cause the calculation of the expected distortion of the current slice to depend on its previous slices. Therefore, we can assume that the current slice depends on its previous z slices (z ≥ 0). Then, given the current decision vectors, the selection of the next decision vector is independent of the selection of the previous decision vectors, which makes the future step of the optimization process independent of its past steps,

44. Forming the foundation of dynamic programming. Thus, the problem can be converted into and solved as a well-known problem of finding the shortest path in a weighted directed acyclic graph (DAG) [11]. In this way the optimization problem is efficiently solved [12]. System Experiments In the experiments video coding is performed by using the H.264/AVC JM 12.2 codec, where the gait video is recorded through an ordinary video camera in an indoor environment, as shown in Fig. 3. The frames of the recorded QCIF sequence are coded at the frame rate (Reframe) of 30 frames/s, where each!-frame is followed by 9 P-frames. We set one packet to be one slice (one row of MBs). When a packet is lost during transmission, we use the temporal-replacement error concealment strategy.

45. The motion vector of a missing MB is estimated as the median of motion vectors of the nearest three MBs in the preceding row. If that row is also lost, the estimated motion vector is set to zero. The pixels in the previous frame, pointed to by the estimated motion vector, are used to replace the missing pixels in the current frame. Furthermore, we adopt the Rayleigh channel model to describe signal-to-noise ratio (SNR) ƴ statistically. For channel adaption, we assume the channel is frequency flat, remaining time invariant during a packet, but varying from packet to packet. Therefore, AMC is adjusted on a packet- bypacket basis. Besides, we also adopt perfect channel state information at the receiver, which is fed back to the transmitter without error or latency. For the joint optimization of QP and AMC, we allow QP to range from 1 to 50 and AMC to be chosen from the six available schemes in Fig. 5.

46. We use NS-2 and MATLAB for system experiments. A serial multi-hop network topology is adopted using NS-2. The bandwidth and the propagation delay on .each link are fixedly set to 106 symbols/s and 10 µs, respectively. The expected video quality at the receiver is measured by the average of the peak SNR (PSNR) of the whole video clip. We compare the PSNRs, under the same network conditions, of the reconstructed video sequences at receiver side achieved by using the proposed system to those achieved by using the existing system of non-content-aware analysis, where the video clip is transmitted with fixed QP and AMC scheme. In the experiments, we set QP to 20 and AMC to 3 for the existing system, respectively. The relation between playback deadline Tt and frame rate Rframe meets Eq. 4:

47. On the basis of this, we consider three different playback deadline values, 20 ms , 30 ms , and 40 ms , respectively. The received video quality achieved through the proposed system in the experiments compared with that through the existing system is demonstrated in Fig. 6. We can observe that in the proposed system, the human motion region based on content-aware analysis has 3-5 dB PSNR improvement over the existing system. Meanwhile, the performance gain of the human motion region over the existing system is even larger in the case of 20 ms than that in the other two cases, indicating that the more stringent the

48. single-packet delay deadline, the more PSNR improvement of the human motion region the proposed scheme can achieve. In other words, the proposed framework is extremely suitable for delay-stringent wireless networks. To explain this, in a more stringent environment, the packet is more likely to miss the playback deadline. Therefore, more error concealment will be used to play back the video, resulting in degradation of the user-received video quality. By adopting the proposed scheme, the optimized parameters can be chosen dynamically to try to meet this playback deadline. Accordingly, higher PSNR gain can be achieved. However, in a less stringent environment, more packets can meet the delay bound even without using the proposed optimization scheme, so there is less opportunity for the proposed scheme to take effect.

49. Therefore, the proposed joint optimization system can choose the best coding parameters and the optimal AMC scheme to decrease packet loss rate by dynamically adapting to the varying channel quality, while trying to avoid missing the packet playback deadline. Additionally, in the experiments the content aware analysis helps reduce the amount of video traffic by around 50 percent compared to the existing non-content-aware analysis system.

50. Conclusions In this article we propose an e-healthcare system based on video content analysis and quality-driven content-aware wireless streaming for remote human gait tracking. The proposed scheme can significantly reduce the reliance on traditional marker-based gait data collection facilities, pro- viding a low-cost high-accuracy gait tracking system. A distortion-delay framework has been proposed to optimize the wireless streaming for delay-bounded retrieval of the collected video data, where key system parameters residing in different network layers are jointly optimized in a holistic way to achieve the best user-perceived video quality over wireless environments.

51. Experimental results have demonstrated that the proposed system can provide great convenience and cost effectiveness for fast prognosis and diagnosis of pathological locomotion biorhythm over resource-constrained wireless networks.

52. A Multimedia Telemonitoring Network for Healthcare TELMES project aims to develop a securized multimedia system devoted to medical consultation teleservices. It will be finalized with a pilot system for a regional telecasters network that connects local telecasters, having as support multimedia platforms. This network will enable the implementation of complex medical teleservices ( teleconsulations , telemonitoring, homecare, urgency medicine, etc.) for a broader range of patients and medical professionals, mainly for family doctors and those people living in rural or isolated regions. Thus, a multimedia, scalable network, based on modern IT&C paradigms, will result. It will gather two inter-connected regional telecasters, in Iasi and Pitesti, Romania, each of them also permitting local connections of hospitals,

53. As communications infrastructure, we aim to develop a combined fix mobile-internet (broadband) links. Other possible communication environments will be GSM/GPRS/3G and radio waves. The electrocardiogram (ECG) acquisition, internet transmission and local analysis, using embedded technologies, was already successfully done for patients’ telemonitoring. Diagnostic and treatment centers, as well as local networks of family doctors, patients, even educational entities.

54. TELMES project aims to develop a securized multimedia system devoted to medical. Keywords-Healthcare, telemedicine, telemonitoring, ECG analysis. In spite of decreased mortality, coronary artery disease still remains the leading cause of death almost all over the world. The existence of silent myocardial ischemia emphasizes the need for monitoring of the asymptotic patient. Extended patient monitoring during normal activity has become increasingly important as a standard preventive cardiologic procedure for detection of cardiac arrhythmias, transient ischemic episodes and silent myocardial ischemia.

55. Existing “halter” devices mostly record "24-hour activity" and then perform off-line record analysis, so they are not real-time. The task may also be achieved by telemedicine (enabling medical information-exchange as the support to distant. This work was supported by a grant from the Romanian Ministry of Education and Research, within CEEX program (www.mct-excelenta.ro), contract No. 604/645/21.10.2005. H. N. Costin , C. Rotariu , and B. Dionisie are with the Faculty of Medical Bioengineering, Univ. of Medicine and Pharmacy, Iasi, Romania, 700098, str. Universitati 16, (phone: [phone]; fax: [phone]; e-mail: [email protected] ).

56. S. Puscoci is with the National Institute of Telecommunications, 062203, B- dul Precise Nr.6, sector 6, Bucharest, Romania. (e-mail: [email protected]). M. C. Compose is with the “Stefan cell Mare” National College, Suceava, Romania, (e-mail: [email protected]). (decision-making) and telemonitoring (enabling simultaneous distant-monitoring of a patient and his vital functions), both having many advantages over traditional practice. A telemonitoring network (Fig. 1) devoted to medical teleservices, will enable the implementation of complex medical teleservices for a broader range of patients and medical professionals, mainly for family doctors and those people living in rural or isolated regions.

57. Doctors can receive information that has a longer time span than a patient's normal stay in a hospital and this information has great long-term effects on home health care, including reduced expenses for health care. Physicians also have more accessibility to experts, allowing the physician to obtain information on diseases and provide the best health care available. Moreover, patients can thus save time, money and comfort. As for patient monitoring, we propose the development of a flexible environment based on an acquisition module and an embedded system for real-time bio signals processing and transmission through Internet, GPRS/3G (mobile telephony) or radio networks already existing in each Romanian county.

58. I. TELEMONITORING MODULE. A. General Structure. Our patient telemonitoring module is based on an ECG / bio signal acquisition module and an embedded system, for real-time signal processing and transmission through Internet (Fig. 2). It is built by using custom developed hardware, open source and application software. For instance, the monitoring device could be used either for acquisition of anomalous ECG sequences (e.g. with arrhythmic events, ST segment deviation etc.) ,

59. and storing to a compact flash memory, as a warning device during normal activity, or an exercise stress test. The heart of the module is an 8–bit microcontroller (uPSD3234A from ST Microelectronics). It has an 8032 compatible UCP capable of being clocked up to 40 MHz. The μPSD has a memory structure that includes two independent Flash memory arrays, main (256 Kb) and secondary (32 Kb), capable of read-while-write operation. It also contains a large SRAM memory (8 Kb) on chip, with battery back-up option for RTOS and communication buffers. The other features of the μPSD include: communication interfaces such as USB v1.1 Low Speed (1.5Mbit/s), I2C Master/Slave controller running up to 833 kHz, SPI Master controller, two UARTs with independent baud rate, IrDA Protocol up to 115 k baud, 4channel 8-bit A/D Converter and 5 PWM channels.

60. The Patient module is connected to a medical device such as Wrist Clinic™ or Monoclinic™ via the USB connector and receives the data. An external USB camera can be connected to a secondary USB connector in order to receive images from the patient. Data can be stored temporarily on the μPSD internal memory or in a Data Memory which is an external memory chip capable to store up to 32Kb. The patient module offers the possibility to make an audio connection with the patient by using the microphone and the speaker connected to μPSD through an external A/D and D/A converters. The ETHERNET controller is RTL8019AS, one of the modern implementations of the NE2000 standard.

61. It integrates 16 Kbytes of SRAM, modulator and demodulator for the physical interface, Ethernet protocol controller, memory interface, and many other functions. The data collected from the Medical devices, waveforms and parameters, can be displayed by using a popular graphical module (128 x 64 pixels) before sending them to a Telemedical Centre by using the Internet connection or by using the GSM Modem. A. Software. The software working on the Patient Module collects the data from medical devices and video/audio sources, computes the parameters and displays them on the LCD Display activates the alarms and sends the results to the Telemedical Centre.

62. The device has as main features: real-time ECG / bio signal acquisition and processing, executes the operator’s commands, monitors the system’s overall performance, acts in emergency situations, and aids the diagnostic. To make a simple software implementation, we choose to use the standard TCP/IP network protocol as the link provider, a scalable and economically feasible tool. For DAQ applications in real-time, such as ours, one must use real time (RT), multitasking operating systems. A modern and economic solution is to choose an open source (free) RTOS, such as RT-Linux. It is comprised of a small RT kernel which runs: (i) a C/C++ RT process at top priority, and (ii) the standard Linux kernel as a fully preempt able low priority task. High speed (low interrupt latency)

63. and predictable timing are achieved by limiting the RT process to functions that are essential to real time. B.ECG Acquisition and Processing. The most important ECG phases for morphological analysis are [1] : • P-wave (representing contraction of the atria); • QRS complex (representing contraction of the ventricles); • T-wave (representing the recovery of the ventricles). Typical ECG processing algorithms consist of the following steps:

64. a) Initialization - used to determine initial signal and timing thresholds, positive/negative peak determination, automatic gain control, etc. b) Filtering - this is performed first as analog filter on ECG amplifier board, and then as digital filter on acquisition board. In addition, a 50 Hz notch filter is used to reduce power line interference. c) QRS complex detection - reliable detection of R-peak is crucial for morphological analysis [3]. d) Baseline correction - compensates for low-frequency ECG baseline drift. e) ST segment detection [6].

65. C. Detection of QRS Complexes An adaptive thresholding technique with search back serves as the primary method for QRS detection. The thresholds are based on the most recently detected signal and noise levels to react to changes in the patient’s heart rate, as well as to signal and noise levels. The QRS complex is the most significant feature in the ECG signal. Being characterized by sharp slopes, its duration is about 70 – 130 msec and its energy spectrum is mostly between 1 and 40 Hz. The input of the QRS detector is the digital ECG signal, sampled at 250 Hz and quantized with 12 bits/sample by A/D converter. The outputs are the limits of the QRS complex ( QRSon and QRSoff ), the location of the R wave, and location of the QRS peaks and notches (if they exist) of every beat (complex) [4], [5].

66. The QRS detection algorithm consists of three steps: (1) coarse QRS limits determination; (2) peaks and notches determination, and (3) exact limits determination. D. ST Segment Analysis. The ST-segment begins 40 msec after the R-peak in the event the heart rate is more than 100 bpm , or 60 milliseconds after the R-peak otherwise. ST-segment has normally a predefined length of 160 milliseconds. The normal Segment template is constructed for each patient as the average of the first ten normal ST-segments. Baseline drift is compensated according to the slope between the isoelectric levels of the two beats. Standard annotated databases, such as the European ST-T Database and the MIT/BIH Arrhythmia Database, provide means for algorithm evaluation. In order to compute ST-segment length, a T wave detector must be implemented [2 ].

67. E. Data Compression and Error Rate. Experiments revealed the necessity for data compression, in order to make a real-time ECG transmission. We used Linux zip programmed, that yields about 2:1 average compression ratio by means of Lempel Zip algorithm. Table I presents results obtained for a resolution of 12 bits/sample and 250 Hz sampling rate, with 3 leads ECG. In this way, only 6 KB/s bit rate is enough for a real-time 3 leads ECG transmission!

68. The quality of data transmission was evaluated by computing PRD (percentage rate of distortion, a kind of root mean square), according to formula (1) below. Table II shows a PRD level under 10%, a value accepted by clinicians for expressing a correct diagnosis.

69. II. RESULTS AND DISCUSSION. A . The whole telemonitoring system acts as a client-server application. The server module includes: a database server (using MySQL and open sources for server procedures, tables, restrictions coming from “client” application); an administration/control module that supervises general dataflow; an access/security module; a parameters configuration module a.s.o . Also, it uses HTML and HTTP to send most up to date information on heart care to clients.

70. B . The client module comprises the software working on the expert's computer. It is implemented by using Java applets and has the following facilities: GUI (Graphic User Interface) for ECG monitoring (Fig. 3); displays the patient’s ECG in real-time and the extracted ECG segments data; communicates the experts' commands (e.g. remote selection of the ECG lead) and medical decisions to the physician/patient. Also, some off – line processing algorithms are implemented, such as: advanced filtering; morphologic ECG analysis (intervals, amplitudes, electrical axes), average complexes with measurement reference markings; heart rate variability analysis, etc.

71. We designed and prototyped the monitoring unit for acquisition and real-time ECG processing, the software implementation for the Internet connectivity (the embedded TCP/IP subsystem), and the software for displaying ECG information on the medical doctor’s computer. The average reconstruction error of the ECG signal is about 4.6%. We also tested various algorithms for morphologic ECG analysis with good results on MIT/BIH Arrhythmia Database (Table III). The utility of our system is as follows . 1.The monitoring. Medical monitoring – to watch the clinical / preclinical parameters of a patient in order: (a) to decide upon a oncological, etiological or prognostic diagnosis and to make a treatment decision, or (b) to assess the efficiency / to correct a treatment plan.

72. 2.Pre-diagnostic monitoring. It is used in order to establish a oncological, etiological or prognostic diagnosis, in the case of discrete, infra-clinical, atypical manifestations of diseases or for obvious signs in case of stress tests / daily conditions. A single parameter in conditions of interrupting the medication is motorized. E.g.: Halter monitoring of ECG in arrhythmias, arterial pressure in border hypertension syndrome. 3. After-diagnostic monitoring (I ). (a) Vital functions monitoring: cardio-vascular, respiratory or vital nervous centers functions in case of : - shock syndrome (cardiac, hypovolemic, politraumatic , septic emic, anaphylactic, post-combustion) ;

73. - comas (traumatic, metabolic – hepatic, uremia, diabetic); (b) only in emergency hospital services, in situations that needs a rapid therapeutic reaction; (c) multi-parameters monitoring devices with alarm systems released by pathological patterns or critical values of: heart or respiratory rate, oxygen saturation, central arterial or venous pressure. 4. After-diagnostic monitoring (II ). Treatment monitoring – posology: - establishes the dose or the efficiency a drug; - establishes the circadian rhythm of maximum effect of a drug; - establishes the secondary effects or adverse effects, or effects of association the drugs; - it is monitor zed the improving of functional parameter under treatment and it make the decision to change the drug or to modify the dose

74. ; - e.g., respiratory flows MEF50, FEV1 under β-agonist treatment in Bronchial Asthma; arterial tension after ant hypertension drugs. 5. After-diagnostic monitoring (III ). (a) Monitoring the recovery and the maintenance of remaining functions: - cardiac, respiratory, (Insufficiency Syndromes) or - locomotion (lack of force / articular mobility – dynamometry, goniometry. (b) Monitoring the recovery with specific functional stress testing: cycloergometer treadmill.

75. 6. Special monitoring. - Telemetric monitoring for– Psychiatric patients (dementia with dromomania); - Persons in extreme environment with physical stresses (deserts, under water, extra-terrestrial, calamities); - Sportsmen training; - Periodical monitoring the health status. (c) Alarm feedback mechanism to watch the level of effort and risk of decompensating.

76. III. SUMMARY AND CONCLUSION. Real time personal ECG monitoring, as an important application of telemonitoring system, requires devices with high peak performance and low power consumption. High performance of RT-Linux development environment allows high speed multitasking procedures and real time signal processing. The proposed system could be used as a warning system (Halter-type) for monitoring of arrhythmia or ischemia during normal activity or physical exercise. In addition to monitoring of physiological signals, we plan to use the proposed environment for development of a high performance user interface. New user inputs, including correlates of the user's physiological and emotional states could significantly improve human-computer interface and interaction.

77. Many algorithms for ECG analysis have already been tested with very good results. Moreover, our monitoring system is general enough to enable a wide range of bio signals monitoring and analysis, e.g. ECG, EEG, EMG a.s.o . ECG tele -monitoring of a patient in real time, according to our project, has as main feature the analysis and transmission of the patients’ bio-signals through the Internet, so that experts in cardiology could make the right diagnostic. So, by using the existing web-based and embedded technologies, the quality of medical decision in tele -healthcare and emergency medical services systems can be significantly improved.

78. Multimedia-Based Healthcare. THE rapid advances in electronic devices, digital imaging, information technology, computer systems, and networks in recent years have stimulated the explosive growth of multimedia computing with diverse applications to different areas including medical service and healthcare. Equipped with various multimedia tools, techniques, and services, computerized healthcare is emerging as an ever-increasing important multidisciplinary area which offers tremendous opportunities and excellent facilities to doctors, healthcare professionals, and other eligible users to enhance performance by fully utilizing the rich health-related multimedia data for effective decision making.

79. The evolution of this new combination of computer based and health-oriented trends will be beneficial not only for computer-aided medical diagnosis and therapy, but also for modern telemedicine and promising e-health applications. Although the current achievements are exciting and the results can be powerful, it remains a challenging task to manage the diversity of health-related multimedia data on an open heterogeneous landscape (multi-modality, big volume, mobility, time series) efficiently, accurately, reliably, and cost-effectively. On the technology front, the use of computing power to handle computation intensive analysis of multi-scale, multi-modal, heterogeneous, and time-variant health related data has led to a new path which encompasses innovations in disease diagnosis, prognosis, customized drug development, and medical monitoring.

80. However, a significant barrier to high standard healthcare services is caused by the lack of powerful methods to handle three fundamental issues involved: 1) data management of time series of health-related record swhich were collected at different time, stored in different formats, and represented by different models; 2) feature selection and fusion of vast and vital medical multimedia information for comprehensive data analysis; and 3) classification and decision support scheme for convenient, reliable, efficient, and cost effective health assessment. It is highly desired to develop new and effective multimedia computing methods for health data infrastructure by integrating multiple, distributed, heterogeneous data sources for high performance with consistency, objectiveness scalability, and cost-effectiveness.

81. This Special Issue on the topic of multimedia-based healthcare presents the latest findings and the state-of-art methodologies for the cross-area research on computerized healthcare, whichisbeyondthecoverageoftheconventionalmedicalimaging.Weselectedsevenpaperswhichareconcernedwithnotonly the fundamental issues on data analytics in the rich context of health related multimedia data, but also the crucial components such as health data representation, management, analysis, classification, and machine learning technologies. We hope that the introduction of these seven papers will provide the insight into the current advances, challenges, and trends. These papers also offer the state-of-art research methodologies, technologies and services with exciting findings and applications.

82. The first three articles are on disease detection and classification by machine learning for different applications. The first article entitled “Classification-based record linkage with pseudonym zed data for epidemiological cancer registries,” by Y. Siegert, X. Jiang, V. Krieg, and S. Bartholomew's, presents a new approach to monitor, evaluate, and predict the development of cancer. The objective of this research is to explore machine learning methods for cancer registries. The main contribution of the work is the proposal of a new method to encode the “pseudonym zed” data for feature extraction and incorporate three classifiers (neural network, support vector machines, and decision tree) for classification-based record linkage to achieve robustness in cancer registries.

83. The comprehensive data analysis of the experimental results demonstrates the promising potential of the proposed approach in real practice. The second article, “ConfidentCare: A clinical decision support system for personalized breast cancer screening,” by A. M. Alaa , K. H. Moon, W. Hsu, and M. van der Schar , is concerned with the development of a computer-aided clinical decision support system with learning capacity from the electronic health record (HER) for personalized breast cancer screening. Most of the existing screening systems are based on a “one-size-fits-all” approach adopted by current clinical practice guidelines (CPGs). Such systems can offer only a general analysis according to the fixed pre-defined rules for all of patients without consideration of the differences in each individual.

84. By contrast, ConfidentCare is developed based on a new theoretical framework for supervised learning, where the space of patients’ features are decomposed into clusters and the learning process is guided by cost-effective and accurate personalized screening policies to ensure a high confidence level of performance bounds for every cluster of patients. More specifically, ConfidentCare operates by computing clusters of patients with Similar features, then learning the “best "screening procedure for each lust erusingasu prevised learning algorithm. The experimental results on practical cases show that ConfidentCare outperforms the current CPGs in terms of cost and efficiency. The third article, authored by Q. Fang and Y. Zhou, is entitled “Kernel combined sparse representation for disease recognition.” The work reported is on the recognition of disease by classification of medical images.

85. One of the key issues is to develop effective classifiers that can be used to detect and categorize diseases. Although sparse representation-based classification (SRC) has been widely used for image classification with good performance, the method can be improved further by considering the correlation structure of the prototype to enhance the performance. The third article presents the combined sparse representation (CSR) classifier to find the relationship between the test sample and the training samples. In addition, the kernel combined sparse representation (KCSR) classifier is adopted to map the original feature to a nonlinear high dimensional feature such that it can obtain the nonlinear information for classification.

86. In order to increase the information of the training samples and the corresponding class mean sample, The proposed classifiers have been evaluated by extensive experiments on several well-known databases including the EXACT09 database, Emphysema-CT database, mini-MIAS database, WBC database, and HD-PECTF database. The experimental results demonstrate that the proposed classifiers achieve better recognition rates than many of the popular methods including sparse representation-based classification and collaborative representation-based classification. The fourth selected article, “Audiovisual spatial-audio analysis by means of sound localization and imaging: A multimedia healthcare framework in abdominal sound mapping” by C. A.

87. Dimoulas, illustrates the trend and significance of multimodality data fusion for health monitoring. The novelty of the method presented in this paper is to apply sound-field localization and visualization techniques for audiovisual spatial audio analysis, content description and management. The focus of the proposed method is topographic analysis and mapping of gastro-intestinal motility (GIM) through multichannel recording of abdominal sounds (AS). Unlike the existing methods which rely on GIM physiology patterns to diagnose specific abnormalities or diseases, a generic framework based on spatial audio analysis is introduced, where multimodal monitoring of GIM and other psycho-physiological parameters can be com binned.

88. For the study of human digestive patterns and their relation to other factors (i.e. subjects’ medical history, nutrition, medication, psychological state, and others). The proposed analysis and management automation can be generally adopted in human bioacoustics, as well as in other spatial-audio applications such as audiovisual surveillance, acoustic scene analysis, soundscape semantics, voice detection, and localization. Mobile health, also referred to as Health, has emerged as an appealing practice of medicine and public health by using mobile devices supported by mobile telecommunication and multimedia technologies. Equipped with technology integration within the health sector, nowadays Health is posed with great potential on different aspects such as offering a better health communication channel to the public for the education and awareness of healthy

89. life style, improving decision making by health professionals to enhance healthcare service by providing diverse medical and health information to facilitate support for diagnosis, treatment, disease monitoring, remote data collection, and epidemic outbreak tracking. The pair of the fifth and sixth articles represent two case studies of Health. The fifth article, “Tensor manifold discriminant projections for acceleration-based human activity recognition” by Y. Guo , D. Tao, J. Cheng, A. Dougherty, Y. Li, K. Yue , and B. Zhang, presents the studies on human activity recognition using wearable accelerometers for health monitoring. The third major contribution of the reported work is the proposal of a new tensor-based feature selection method named “tensor manifold discriminant projections (TMDP)” for classification.

90. The novelty of the propose algorithm is rotated around three major steps during the following operation: 1) applying an optimization criterion that can directly process the tensor spectral analysis problem so as to reduce the computational cost, 2) extracting local rank information by finding a tensor subspace that preserves the rank order information of the within-class input samples, and 3) extracting discriminant information by maximizing the sum of distances between every sample and their interclass sample mean. Experiments on the naturalistic mobile devices-based human activity (NMHA) 2.0 dataset are performed with benchmarking with other methods to demonstrate the effectiveness and robustness of the proposed TMDP approach.

91. The sixth article, authored by S. Cical’o , M.Mazzotti, S.Moretti, V. Tralli , and M.Chiani , addresses another important application of Health: “Multiple video delivery in m-Health emergency applications.” The challenging issue is how to deliver multiple video streams smoothly from an ambulance to a remote hospital over a single bandwidth-limited wireless access channel rather than a large band width. The author Softhis article proposed con text-aware approach with three major objectives: 1) to maximize the QoE for the final user, 2) to efficiently manage distributed hardware resources,and3)to optimize the usage of radio resources for video transmissions. To achieve the objectives, a new optimization framework is introduced, which can transmit data over a single bandwidth-limited wireless access channel by automatically select and adapt the best video streams for data transmission.

92. The major contribution of the proposed approach is as follows:1)a camera ranking technique for the VSN deployed in the emergency area, and 2) a novel solution for the transmission of the videos from the emergency area, based on the joint selection, adaptation and aggregation of the streams directly performed at the application layer of the processing equipment inside the ambulance. Numerical simulations considering a realistic emergency scenario with LTE-Advanced connectivity show the feasibility of the proposed approach. Given the sensitive nature of an individual’s medical information, integrity, security, privacy, and confidentiality are critical issues that must be clearly and effectively addressed by Health applications. X. Yuan, X. Wang, C. Wang, J. Weng , and K.

93. Ran presented a secure cloud-based framework for healthcare monitoring systems with privacy guarantees in the seventh selected article “Enabling secure and fast indexing for privacy-assured healthcare monitoring via compressive sensing.” The proposed framework is a result from an integration of cross-area technical advancement in different domains such as compressive sensing with unified data acquisition and compression, practical encrypted search techniques, high-performance content based indexing, and novel on current program Ming algorithms. The design of system architecture and indexing structure enhance secure and very fast indexing of continuously generated compressed data samples for health monitoring, and enable high-performance encrypted search over compressed samples.

94. Experimental evaluation on Amazon Cloud demonstrates the promising potential of the proposed approach . Today, computerized healthcare is furnished with various applications and encouraging achievements. The rapid advance in multimedia computing offers an appealing synergy to manage medical big data with automated, objective, and scalable measurement reliably, efficiently, and cost-effectively. In this special issue, we aim to strive and foster the state-of-art research on a general platform of multimedia computing for the cross-area healthcare applications. We hope that the selected seven articles form the basis for further development of multimedia-based healthcare applications.

95. The guest editor team would like to take this opportunity to thank all of the people who helped and supported us indifferent ways during the preparation of this Special Issue: The recommendation and support from the TCMC selection committee chaired by Prof. S.-C. Chen for our initial proposal on this feature topic of “multimedia-based healthcare,” the hard work and contribution from all of the authors of submitted manuscripts, thetimeandeffortmadebyallofthereviewers,andtheconsideratearrangementandcoordinationbytheTMMeditorialteam.

96. The proposed network architecture is designed as four layers: perception layer, network layer, cloud computing layer, and application layer. In the network, the integration of TCP/IP and Zigbee in the coordinator devices is utilized. Consequently, WBAN coordinators can compatibility inter-operate with various local networks such as WiFi and LTE network to support high mobility of users. Besides, we integrate Content Centric Networking (CCN) with our proposed architecture to improve the ability of the WBAN coordinator. Thus, it can support uninterrupted media healthcare content delivery. In addition, adaptive streaming technique was also utilized to reduce packet loss. Various simulations were conducted using OPNET simulator to show the feasibility of the proposed architecture in terms of transmitting a huge amount of media healthcare data in real-time under traditional IP-based network.

97. The latest medical information, when expressed in understandable terms, can empower patients to takecharge of their health, take appropriate preventive measures and make more informed choices regarding their treatment. In this paper, we report on the ongoing development of PERSIVAL ( PErsonalized Retrieval and Summarization of Image, Video And Language), a system designed to provide personalized access to a distributed digital library of medical literature and consumer health information. Our goal for PERSIVAL is to tailor search, presentation, and summarization of online multimedia information to the end user, whether patient or healthcare provider. PERSIVAL utilizes the online patient records at New York Presbyterian Hospital [7] as a sophisticated, pre-existing user model that can aid in predicting user interests.

98. References. 1. R. Davis, "Clinical Gait Analysis," IEEE Engineering in Medicine and Biology Mag., vol. 7, Sep. 1988, pp. 35-40. 2. D. Rodriguez-Silva eta/., "Human Motion Tracking and Gait Analysis: A Brief Review of Current Sensing sys- tems and Integration with Intelligent Environments," IEEE VEC/MS, July 2008, pp. 166-71. 3. M. Piccardi , "Background Subtraction Techniques: A Review," Systems, Man and Cybernetics, vol. 4, Oct 2004, pp. 3099-104. 4. C. Stauffer and Eric, "Learning Patterns of Activity Using Real- Time Tracking," IEEE Trans. Commun ., vol. 46, May 1998, pp. 595-602. 5. Q. Jackson and D. Land grebe, "Adaptive Bayesian Con- textual Classification Based on Markov Random Fields," IEEE Trans. Geoscience and Remote Sensing, vol. 11, Nov. 2002, pp. 2454- 63 .

99. 6. R. Kinderman and J. L. Snell, "Markov Random fields and Their Applications," Amer. Math. Soc., vol. 1, 1980, pp. 1-142. 7. D.Chai and K. N. Ngan , "Face Segmentation Using Skin-Color Map in Videophone Applications," IEEE Trans. Circuits Sys. Video Tech., vol. 9, June 1999, pp. 551-64. 8. "Draft ITU-T Recommendation and Final Draft Interna - tional Standard of Joint Video Specification (ITU-T Rec. H.2641SO/IEC 1449610 AVC)." Available from ft p:1/ftp.imtc-1es.org/jvt-experts/2 00303 Pattaya /JVT- G50r1 .zip, May 2003. 9. R. Zhang, S. L. Regunathan , and K. Rose, "Video Coding with Optimal Inter/Intra-Mode Switching for Packet Loss Resilience," IEEE JSAC, vol. 18, no. 6, June 2000, pp. 966-76.

100. 10. Avgerinakis , K., Briassouli , A., & Kompatsiaris , I. (2013). Activity detection and recognition of daily living events. In  1st ACM MM workshop on multimedia indexing and information retrieval for healthcare (MIIRH), in conjunction with ACM MM 2013 . Barcelona, Spain, pp~3–10, October 2013. Google Scholar 11. Helal , A., Mokhtari , M., & Abdulrazak , B. (2007).  The engineering handbook on smart technology for aging, disability and independence . Hoboken: John Wiley and Sons. Google Scholar 12. Mulin , E., Joumier , V., Leroi , I., Lee, J. H., Piano, J., Bordone , N., et al. (2012). Functional dementia assessment using a video monitoring system: Proof of concept.  Gerontechnology , 10 (4), 244–247. Google Scholar 13. Chen, S. W., Lin, S. H., Liao, L. D., Lai, H. Y., Pei, Y. C., Kuo , T. S., et al. (2011). Quantification and recognition of Parkinsonian gait from monocular video imaging using kernel-based principal component analysis.  Biomedical Engineering Online, 10 , 99. doi :  10.1186/1475-925X-10-99 .

101. 14. P. Laguna, et al., "New algorithm for QT interval analysis in 24- hours halter ECG: performance and applications", Med. Biol. Eng. Compute., Vol.28, Jan. 1990, pp. 67-73. 15. J. Pan, W.J. Tompkins, "A real-time QRS detection algorithm", IEEE Transactions on Biomedical Engineering, Vol. 32, No. 3, March 1985, pp. 230-236. 16. D.L. Rollins, et al., “A telemetry system for the study of spontaneous cardiac arrhythmias”, IEEE Transactions on Biomedical Engineering, Vol. 47, No. 7, July 2000, pp. 887-892. 17. A. Ruha , et al., "A real-time microprocessor QRS detector system with a 1-ms timing accuracy for the measurement of ambulatory HRV", IEEE Trans. on Biomedical Engineering, 44, No. 3, 1997, pp. 159-167. 18. V. Subramanian, “Clinical and research applications of ambulatory Halter ST-segment and heart rate monitoring”, The American Journal of Cardiology, Vol. 58, No. 4, 1986, pp. 11B- 20B

102. 19. European Commission, Information Society Technologies Directorate General, “Resource Book of e-Health Projects – 6th R&D Framework Program me, 2002-2006”. 20. European Commission, Information Society Technologies Directorate - General, “Applications Relating to Health– 5th R&D Framework Program me, 1998-2002