Information technologies: opportunities and ...

Download PDF
8 downloads 165 Views 322KB Size Report
Comment
using information technologies to deliver personal healthcare. Keywords: ... Dat Tran received his BSc and MSc degrees from University of Science,. Vietnam ..... The moods of the patients were subject the environments after the MQ practice.
Int. J. Healthcare Technology and Management, Vol. 13, Nos. 5/6, 2012

Information technologies: opportunities and challenges in personal healthcare systems Wanli Ma* and Dat Tran Faculty of Information Sciences and Engineering, University of Canberra, ACT 2601, Australia E-mail: [email protected] E-mail: [email protected] *Corresponding author

Hong Lin Department of Computer Science, University of Houston, Houston, TX 77002, USA E-mail: [email protected]

Shang-Ming Zhou Health Information Research Unit, College of Medicine, Swansea University, Swansea, SA2 8PP, Wales, UK E-mail: [email protected]

Byeongsang Oh Sydney Medical School, University of Sydney, NSW 2006, Australia and Dana-Farber Cancer Institute, Harvard Medical School, Boston, MA 02115, USA E-mail: [email protected]

Gordon Waddington Faculty of Health, University of Canberra, ACT 2601, Australia E-mail: [email protected]

Copyright © 2012 Inderscience Enterprises Ltd.

345

346

W. Ma et al.

Dharmendra Sharma and Mohammad A. Rahman Faculty of Information Sciences and Engineering, University of Canberra, ACT 2601, Australia E-mail: [email protected] E-mail: [email protected]

Ongard Sirisaengtaksin Department of Computer Science, University of Houston, Houston, TX 77002, USA E-mail: [email protected]

Jennie Scarvell and Tim McGrath Faculty of Health, University of Canberra, ACT 2601, Australia E-mail: [email protected] E-mail: [email protected]

Dat Huynh Faculty of Information Sciences and Engineering, University of Canberra, ACT 2601, Australia E-mail: [email protected] Abstract: The well-being of a person consists of two aspects, the physical and the psychological. Technological development makes it possible to massproduce cheap sensors for personal use. The data collected provide objective and comprehensive personal health information. In this paper, we report our preliminary findings in applying modern information technology to personal healthcare systems. We are constructing a brain activity level model by using EEG signals to objectively measure the effectiveness of meditation, detect mental fatigue and boredom, and comprehend human emotions. Also we have used accelerometer and GPS data to assess sports performance and training enhancement, lower limb injury prevention and recovery monitoring, and falls prevention for aged people. Then we exploit the potential of Kinect devices in monitoring the movements of aged persons in their houses to prevent falls. Finally, we point out some remaining challenges and possible opportunities in using information technologies to deliver personal healthcare. Keywords: data mining; machine learning; eHealth; meditation; medical qigong; MQ; body movement; postural sway; electroencephalographic; EEG; accelerometer; global positioning system; GPS; Kinect.

Information technologies Reference to this paper should be made as follows: Ma, W., Tran, D., Lin, H., Zhou, S-M., Oh, B., Waddington, G., Sharma, D., Rahman, M.A., Sirisaengtaksin, O., Scarvell, J., McGrath, T. and Huynh, D. (2012) ‘Information technologies: opportunities and challenges in personal healthcare systems’, Int. J. Healthcare Technology and Management, Vol. 13, Nos. 5/6, pp.345–362. Biographical notes: Wanli Ma received his PhD in Computer Science and Technology from the Australian National University, Canberra, Australia, in 2001. He has eight year’s first hand experience in running IT infrastructure, IT security operations, and digital forensic investigation before joining the Faculty of Information Sciences and Engineering in January 2004. His research areas are of application nature and mainly concentrate on the application of machine learning and artificial intelligence techniques into cyber security, anomaly (intrusion, spam email, and virus, etc.) discovery and detection, internet identity management, digital evidence gathering and analysis, and human biological signals. Dat Tran received his BSc and MSc degrees from University of Science, Vietnam, in 1984 and 1994, respectively. He received his PhD in Information Sciences and Engineering from University of Canberra, Australia in 1996 and 2001, respectively. Currently, he is an Associate Professor at Faculty of Information Sciences and Engineering, University of Canberra. He was an intern for the Pen Technologies Group, IBM T.J. Watson Research, New York, from June to September 2000. His research interests include brain-computer interfaces, robotics, biometric authentication, fuzzy modelling, spam email detection, language identification and cell phase recognition. Hong Lin received his PhD in Computer Science in 1997 at the University of Science and Technology of China. Before he joined the University of Houston-Downtown (UHD), he was Post-doctoral Research Associate at Purdue University, Assistant Research Officer at the National Research Council, Canada, and Engineer at Nokia, Inc. He is currently an Associate Professor with UHD. His research interests include parallel/distributed computing, grid computing, multi-agent systems, and neurological computation. He is a Co-supervisor of the Grid Computing Lab at UHD. He is also a senior member of the Association for Computing Machinery (ACM). Shang-Ming Zhou is with Health Information Research Unit, Institute of Life Sciences, College of Medicine, Swansea University, UK. His research interests include health informatics, computational intelligence and data mining for improving public health, (type-1 and type-2) fuzzy logic based uncertainty modelling in health informatics; machine learning with uncertain and imprecise information, pattern recognition, uncertain information aggregation for health decision supports (e.g., via Yager’s OWA operator, type-1 OWA operator, type-2 OWA operator, etc). He has published widely on these topics. Byeongsang Oh is Research Fellow at the Harvard Medical School, Dana-Faber Cancer Institute and Clinical Senior Lecturer at the Sydney Medical School, University of Sydney. He established the Integrative Medicine Services at Mater Hospital, Royal North Shore Hospital (RNSH) and Sydney Adventist Hospital (SAH) in Australia, and specialise in integrative oncology (consulting evidence-based complementary medicine for cancer care). His main research area is integrative medicine for cancer care. He is a member of COSA, ASCO, SIO, AACMA, and CMBA.

347

348

W. Ma et al. Gordon Waddington is Professor of Physiotherapy and Associate Dean Research in the Faculty of Health, University of Canberra. He gained his PhD from the University of Sydney in 2000. He has published more than 60 research papers in peer reviewed journals and is currently an Associate Editor of the Journal of Science and Medicine in Sport and on the review boards of the Australian Journal of Physiotherapy, Experimental Brain Research, British Journal of Sports Medicine, Perceptual and Motor Skills, Physical Therapy in Sport and Physiotherapy and the Grants Review Panel of the Physiotherapy Research Foundation. Dharmendra Sharma is the Dean of the Faculty of Information Sciences and Engineering, University of Canberra. He has been with the University of Canberra since February 2001 where he has worked in a number of academic positions and has held several senior university positions. He has special interest in industry engagement, courses for industry and applied research. His research background is in the areas of artificial intelligence, data mining, predictive modelling, constraint processing, fuzzy reasoning, hybrid systems and their applications to health, education and sports. He has published over 200 research papers in these areas. Mohammad A. Rahman is a part-time student of the Professional Doctorate in Information Sciences at the University of Canberra, Australia. He completed his Master of Information Technology in 2005 and Bachelor of Computer Science and Engineering in 2002. Currently, he is working as a Computer Programmer. Ongard Sirisaengtaksin is a the Co-Director of Grid Computing Lab at the University of Houston-Downtown. He has held his current position of Professor of Computer and Mathematical Sciences at the University of Houston-Downtown since 2000. His research areas of interests include intelligent control systems, multi-agent systems, parallel computing, 3D modelling and visualisation. He currently serves as PI on the NSF CI-TEAM, Minority Serving Institutions – Cyber-Infrastructure Empowerment Coalition, Co-PI on the NSF S-STEM award, Undergraduate/Graduate Student Immersion in STEM, and the NSF MRI, Acquisition of a Computational Cluster Grid for Research and Education in Science and Mathematics. Jennie Scarvell is Physiotherapist and Academic. Her area of interest is musculoskeletal physiotherapy, with a particular interest in rheumatology, juvenile chronic arthritis, orthopaedics and biomechanics. She worked in these fields clinically for many years, before completing her PhD in 2004 entitled Kinematics and Degenerative Change in Ligament Injured Knees and explored the development of osteoarthritis in injured knees. She is a Visiting Fellow at to the Trauma and Orthopaedic Research Unit at Canberra Hospital. Projects include prevention and treatment of fractures, knee motion in injury and in total joint replacement, and medical imaging particularly for three-dimensional modelling. Tim McGrath is a part-time PhD student of the University of Canberra. Dat Huynh is a PhD student of the University of Canberra and also a part-time Teaching Fellow of the Faculty of Information Sciences and Engineering.

Information technologies

1

349

Introduction

The well-being of a person consists of two aspects: physical well-being and psychological well-being (the perception or the feeling of well-being). However, data collection on such a large scale and personal level, and also in real-time life environments on the states of the physical body and mind of a person was not possible in the past. Information technology development makes it possible for the general public to access a wide variety of sensors. For example, smartphones, such as iPhones and Android phones, all come with a range of sensors. Most of them have an ambient light sensor, proximity sensor, global positioning system (GPS), accelerometer, gyroscope, microphone, and camera, depending on configuration. Nowadays, smartphones are almost undetectable to the increasing number of human users. Advances in modern manufacture technologies mean cheap data collecting devices are available in many areas, such as electroencephalography (EEG) headsets, Kinect, and thermal imaging cameras. These sensors together can collect a rich set of data relating to the person who carries the device. The data collected, being so specific to this very individual, are very valuable sources of information for personalised healthcare to this person. The information helps us to understand the well-being of the person and then further offers the opportunity to develop a high quality personal healthcare system for the well-being of the person. An EEG device can record the electric signals from a human scalp. EEG devices used to be only available in professional healthcare institutions for clinical use. The last decade witnessed the development of cheap EEG devices, for example, EPOC from Emotiv (2012) and NeuroSky (2012), and increasing interest in EEG-based brain-computer interfaces (BCI). EEG signals characterise the result of the neuronal activity of a human brain. Naturally, they are used to study and understand human brain activities. In particular, EEG signals indicate that neural patterns of meaning in each brain occur in trajectories of discrete steps, whist the amplitude modulation in EEG wave is the mode of expression of meaning (Freeman, 2000). Zhou et al. (2008) have proposed some novel features for EEG signals to be used in BCI systems for classification of left and right hand motor imagery. The experimental results have shown that based on the proposed features, the classifiers using linear discriminate analysis, support vector machines (SVM) and neural networks can achieve better classification performance than the BCI-competition 2003 winner on the same dataset, in terms of the criteria of either mutual information or misclassification rate. Dressler et al. (2004) studied the effect of anaesthetics on the level of sedation. Hamadicharef et al. (2009) suggested the use of EEG signals to measure the attention level, which can be used to detect attention deficit hyperactivity disorder (ADHD). Lin et al. (2010) studied the change of human emotion while listening to music through EEG signals. With the understanding of brain states, we propose to develop an objective measurement to gauge the effectiveness of meditation activities. It has long been believed that meditation contributes positively to the well-being of human minds and bodies. Clinical trials have also confirmed, to some extent, that meditation improves the quality of life (QOL) of terminally ill patients (Oh et al., 2008, 2010). However, individuals do have very different experiences of meditation outcomes. To study the degree of meditation benefits, an objective measurement of the effectiveness of the meditation has to be established. We believe that there is a direct relationship between the level of brain

350

W. Ma et al.

activity and the effectiveness of meditation, and thus the level of brain activity can be used as an objective measurement. The measurement also helps improve the teaching and practice of meditation. It will lead to more effective teaching, better practice outcomes, and wider acceptance (and benefits) of meditation in the general population. Furthermore, the measurement can be further used to detect mental fatigue and boredom and comprehend human emotion. On the other hand, around 31% of adults aged 15 and over in the world were insufficiently active in 2008 (men 28% and women 34%), according to the World Health Organization (2012), and approximately 3.2 million deaths each year are attributable to insufficient physical activity. Indeed, physical inactivity has repeatedly been shown to be a major contributor to health problems in affluent societies. Better physical function can improve mental well-being and reduce risk of chronic diseases. GPS and accelerometer data can provide very detailed information about the body movement of the user. Brophy and her group carried out a study by using the GENEA accelerometer to examine the possibility of recording and identifying activity patterns of infants to promote healthier lifestyles (Brophy et al., 2012). Zhang et al. (2010) suggested that the data collected from the accelerometers of smartphones can be used to gain a further insight into the daily activities either in the home or in outdoor environments and also to identify six activities offline (walking, posture transition, gentle motion, standing, sitting and lying). Lane et al. (2010) proposed that the combination of accelerometer data and a stream of location estimates from the GPS can recognise the mode of transportation of a user, such as riding a bike, driving a car, taking a bus, or on subway. For competition sports, body movement data provide detailed analysis of the performance, training, and physical body status of the players. Nowadays, athletes very often wear data collecting devices, for example, SPI Pro X II from GPSports (2012), when in training and competition. The data can also reveal if an athlete is fatigued and thus may prevent injuries. In clinic, collected accelerometer data can be used to measure the time needed for a person to reach a stable postural position, called time-to-stabilisation (TTS), in a single-leg stance test. The measurement is used by physiotherapists to quantify the postural stability of the person. Therefore, these data are valuable in both assessing injuries and monitoring the recovery process. Likewise, accelerometer data can be used to constantly monitor the postural stability of older people to prevent falls, which may have fatal consequences. The use of accelerometer data in these areas is gaining more interest in the research community and in industry. For example, Kamen et al. (1998) and Waarsing et al. (1996) suggested using accelerometer data to assess balance and postural sway, and Yoshida et al. (2006) proposed a method to detect lower limb injuries by using accelerometer data. Falls are a major cause of death in the elderly. It is not the fall itself but the complications caused by the falls that are responsible for mortality. Falls account for over 80% of all injury-related admissions to hospital of people over 65 years. Falls are the leading cause of unintentional injury mortality in these individuals and responsible for appreciable morbidity (Kannus et al., 2006), according to statistics, 70% of accidental deaths in persons over 75 (Andonegi, 2006). Lee and Carlisle suggested the use of accelerometers to detect falls of aged persons (O’Sullivan et al., 2009; Lee and Carlisle, 2011). However, much depends upon the seniors’ willingness to wear the device. Older generations may not be as keen as the younger ones to carry these ‘high technology gadgets’ all the time and regard them un-detachable. In

Information technologies

351

addition, old people tend to stay in their own houses most of the time. This environment makes them even more reluctant to wear extra foreign objects. The Kinect device released by Microsoft recently allows the freedom to track motion without having the person wear sensors (Kinect, 2011). Our research was motivated to use the Kinect device to explore its use in health monitoring of the elderly, due to its popularity (Records, 2011), low cost and free SDK to develop application for non-commercial purposes. Collecting such a rich set of data always raises some privacy concerns (Simkevitz, 2009). Indeed, these data are very personal, and most people do not want them to be shared. However, the use of these data in personal healthcare is not a problem since the data collected are only used for the concerned individual. There is neither sharing nor disclosure of the data to any other parties. There is a possibility of the data being exposed to the others if the device is lost, but proper encryption and access control prevent unauthorised accesses. On the other hand, the devices themselves, being passive, only loyally perform the tasks of collecting and processing data. If used properly, such a system of using technology for personal healthcare actually helps to protect the privacy of the users, since the basic monitoring and advice is done by the device and no human practitioners need to be involved. In this paper, we report our joint efforts of bringing in computer scientists, health professionals, and physiotherapists together to experiment with the applications of information technology to personal healthcare. We are conducting experiments using EEG headsets, accelerometers, Kinect devices and thermal imaging cameras. We have carried out a series of preliminary experiments, gained much insight in connecting the information technology with the well-being of individuals, and formed a strong position on the tasks to make effective and efficient personal care by using the technologies. The rest of the paper is organised as follows. Section 2 discusses the use of EEG signals in measuring the effectiveness of meditation. Section 3 reports our study of body movement through accelerometer and GPS data, and Section 4 demonstrates the potential of using Kinect devices in aged care. In Section 5, we discuss the opportunities and challenges in applying the information technologies to personal healthcare system. We summarise the paper in Section 6.

2

Meditation and MQ for the QOL

Meditation is a way to achieve the attainment of awakening and enlightenment. It deemphasises any rational creeds or theoretical reasoning to free or calm one’s mind. According to Encyclopedia Britannica Online, Transcendental Meditation (TM) gained popularity in the west during the 1960s (Hölzel, 2011). Medical qigong (MQ) is a type of meditation. One of the authors conducted a series of studies in investigating the practice of MQ in improving the QOL of terminally ill patients (Oh et al., 2008). Eight heterogeneous cancer patients were randomised to undertake MQ once or twice a week for eight weeks, with ten patients randomised to the control group. The intervention group reported improved QOL and reduced symptoms of side effects of cancer treatment after eight weeks. A further study conducted by the same author compared QOL, fatigue, mood, side effect of treatment (nausea, pain, sleep and satisfaction with sexual life) in 162 heterogeneous cancer patients. Seventy-nine were randomised to receive qigong

352

W. Ma et al.

twice a week for ten weeks and 83 were randomised to the control group (Oh et al., 2009). Regression analyses found significant improvements in QOL, and significant reductions in fatigue, mood disturbance, side effects of treatment and inflammation in the intervention group compared to the control group. Such results suggest MQ can improve important health-related outcomes in cancer patients. The limitations of the study include the fact that as the control group were only given usual care, the extra attention given to the intervention group was not controlled for. Further, this study evaluated the effects of MQ on self reported cognitive function and suggests that MQ benefits cancer patients’ self-reported cognitive function (Oh et al., 2012). The past research work was based on patients’ self reported questionnaires after MQ practice. To a great large extent, the answers reflect the feeling of the patients at the time. However, it does suffer from two drawbacks. First, filling the questionnaire is subjective and indirect. It cannot objectively or accurately record the effectiveness of MQ practice. Second, answering questionnaires is posterior. It happened after MQ practice. The moods of the patients were subject the environments after the MQ practice. The environments are of the time of reporting and are different from these when practising. To accurately and objectively record moods when one is practising meditation, we sought a solution which could objectively measure the effectiveness of meditation in real time. We started with a project that aimed to create an application that takes EEG data and exposes it to various analytical techniques so the resultant brain states can be studied and predicted. Although this project is still in its early stages, we anticipate that, upon completion, this software can be used to produce important and dependable conclusions about a given subject’s brain state and correlate that to an identified physical or psychological activity, and ultimately we will be able to build a brain-state model for meditation. The concurrent brain states associated with TM were viewed as something outside of the world of physical measurement and objective evaluation by most scientific communities. Due to the easily obtainable EEG headsets, recording EEG signals can be performed in a large scale. It is therefore possible to build a model for meditation brain state (Lin, 2010). By applying data-mining algorithms that quantify psychological states, we expect to analyse brain states associated with meditation to build models for meditation brain states (Davidson et al., 2003). Comparison of the results obtained from different methods can be performed to fuse different models in order to have a deeper understanding of the central and peripheral nervous systems’ role in attaining different levels of mediation. The practical significance of finding a meditation model includes establishment of the guidance for effective meditation exercises and a methodology for verifying the effectiveness of meditation methods. A scientific meditation model will advance studies in natural computer interfaces, identifying depression and mental illness, detecting fatigue and boredom, and comprehending human emotion, for example. Any advances in these areas will have great social, economic, and technical significance. We have conducted a trial collection of EEG data on a patient who performed MQ and other subjects performing tasks with different levels of brain activity, including attempts to rest the brain. The Emotiv EPOC headsets we used can collect 14 channels of EEG signals. The locations of the contact points on the scalp, called nodes, are demonstrated in Figure 1.

Information technologies Figure 1

Nodes AF3 through AF4 (counter clockwise) (see online version for colours)

Figure 2

Packet 1 of nodes AF3, F7, F3, FC5, F4, FC6, F8, and AF4 (see online version for colours)

353

The dataset analysed so far comes from a study in which a candidate performed various tasks alternately. The individual alternated between idle activity, reading news headlines, and participating in a mathematics exam limited to basic algebra every 60 seconds for 20 minutes. These data include 123,001 samples for all 20 minutes. With 129 samples per packet we have approximately 953 packets per node. We can assume that this leaves approximately 47 packets per node per minute. Then we apply linear regression to the generated scatter-plot, the positive or negative slope correlates to an increase or decrease in brain activity for the entire packet of said node. In the first packet of our dataset, we notice correlations between two sets of four nodes ({AF3 F3 FC5 F7} and {AF4 F8 FC6 F4} respectively). By cross examining node position from Figure 1 with the first packet of each node in Figure 2, we can observe similar scatter-plots that are geometrically symmetric when referencing nodes regarding both left and right hemispheres of the

354

W. Ma et al.

frontal lobe. This tells us that {AF3 F3 FC5 F7} and {AF4 F8 FC6 F4} are sections of the brain that work together when the user is in an idle brain state. Our efforts are currently invested into recognising repeatable patterns throughout the packets. Our developed signal processing algorithms will be used to determine repeatable characteristics in sets of packets that belong to each of the idle, news headline, and mathematics test states. We also have developed a first version of an iOS application for iPad to analyse and visualise collected EEG data. This application can parse and visualise EEG data. It is also possible to extend this app to collect EEG data in real time, using a wireless EEG device. This will enable users to record, visualise, and analyse data in real time.

3

Body movement, physical performance, and human health

Human beings are very much less physically active in modern society than previously. People spend most of their time engaged in sedentary work in offices, driving cars, and watching TV, etc., and spend very little time on physical activities. Understanding the essential importance of physical activity to population health, the authorities spend much effort in public campaigns, yet a majority of the population tend not to make the personal connection between their own physical activity patterns and their health. The question of how people will successfully initiate and maintain an active lifestyle remains challenging. We believe that a personalised fitness assistant with individualised and targeted messages can attract attention and elicit behaviour modification from the individual concerned. The amount of gross body movement is a good indication of the level of physical exercise of the body. By continuously monitoring body movements, we can obtain an accurate measurement of the level of physical exercise of the body. Therefore, we can provide personalised and targeted advice through the system to the individual to adjust the level of physical activity to assist in achieving optimal exercise levels for health. Mobile phones have become ubiquitous and are almost permanently attached to many individuals. Most modern smartphones have inbuilt accelerometers for the purpose of measuring the orientation, i.e., vertical and horizontal positions, of the phone. Carried on a human body, the mobile orientation information obtained through its accelerometer provides a continuous indication of the movements of the body. By applying artificial intelligence, we can study the patterns of the orientation information of the mobile phone to decide the human activities, e.g., jogging, walking, sitting, and standing, etc., and the duration of any activities undertaken. Based on the activities of the day and using intelligent software, appropriate advice to improve physical activity patterns can be tailored to individual needs. The time needed for a person to reach a posturally stable position, called time to stabilisation (TTS), in a single leg stance test is used to measure the postural stability of the person. This measurement has wide applications in clinical practice and also athletic training. The calculation of TTS at present is by recording ground reaction forces in three dimensions on a force platform (Ross and Guskiewicz, 2003; Ross et al., 2005). This practice is very expensive and restrictive, as a force plate is very expensive and has to be set up and calibrated to each very well designed in-house location. Thus, the tests can only be carried out in this environment. Given the ever increasing use of cheap, yet accurate, accelerometers, which are commonly built into smartphones and other handheld

Information technologies

355

devices, it is logical to ask if the data collected from an accelerometer can be used to calculate TTS. Our preliminary collection of data on the patterns of single leg standing TTS patterns is very encouraging. However, we do observe greater discretion of using accelerometer data to calculate TTS. The mounting locations of the accelerometer on the body contribute very much to this greater discretion. The nosiest accelerometer data we collected is when the subject held the accelerometer in his or her hand when performing the jump. In sports, some benchmark tests, for example, single leg standing and Illinois agility test, etc., are routinely used to assess the fitness and performance of the athletes. However, except for just a few parameters, e.g., the duration and speed, which can be accurately measured, to a large extent, the judgement is still based on the observation and opinions of the experts in the field. This approach serves the purpose, but does have its limitations. First, the tests are restrictive. A court has to be set up to perform the tests, and the subjects can only perform the required actions during the test time. Second, the test outcomes, for example, whether a subject has good lower limb symmetry, very much depend on the subjective observation of the experts. Accelerometer and GPS devices open a completely new door and great potential for fitness tests. The accessibility of accelerometer and GPS devices within smartphones means that they can provide continuous data on body movement patterns, beyond the test environment and throughout normal functional activities. Understanding the data enables us to accurately and also objectively measure the dynamics of the subject’s movement. Athletes now often wear dedicated accelerometer and GPS devices, such as SPI Pro X II from GPSports (2012), during their training and competition sessions. Analysing these data helps the coaches to understand the performance and the effectiveness of the training and competition. Furthermore, machine learning algorithms can be used to mine patterns from the collected data, and we also collect and analyse the data in a real time mode. Our future work will focus on trials of recreational cycling and open field country snow skiing. The information technology developed for testing the fitness and agility of athletes can also be used to prevent leg injuries. Very often the injuries are cause by fatigue. An early sign of fatigue development is the higher degree of body sway. When a body starts to lose balance, it suggests that the leg muscles are fatigued and cannot execute sufficient neuromuscular control. If being forced to perform, injuries may follow. The real time movement monitoring system we are developing will be able to warn the subject in advance to avoid serious injuries to leg muscles. The same system can also be used by physiotherapists to monitor the recovery process of patients with leg muscle injuries and then adjust therapeutic programmes accordingly. Yet another potential application of information technology is for aged care. Increasing postural sway may be an indicator of higher falls risk, and an opportunity to take preventative action. We have collected data samples using SPI Pro X II devices on subjects performing full speed straight running (Figure 3). The X axis data are the recording of forward acceleration, the Y axis data are of vertical acceleration, and the Z axis lateral acceleration. The device samples the three dimensions of acceleration at 100 Hz. The peak marker line marks each step of the running cycle by finding the largest Y axis peaks. Two steps form one stride, i.e., from the left foot striking the ground to the left foot striking the ground again, or from the right foot to the right foot. The degree of body balance is calculated for each stride. In this study, we focus on deep analysis of the degree of body sway during the course of full speed straight line running. We also

356

W. Ma et al.

recognised the activities, running, walking, and turning, etc., through their characteristic acceleration patterns. Figure 3

Samples of acceleration patterns (see online version for colours)

Information technologies

4

357

Kinect and its application in healthcare

The Kinect device gets 3D scene information from its 3D depth sensors. This sensor consists of an infrared laser projector combined with a monochrome CMOS sensor, which captures video data in 3D under any ambient light conditions. The device also has a RGB camera and a multi-array microphone for speech recognition. The skeletal tracking tool provided with the Kinect SDK was modified to collect the joint data. The Kinect device records the joints as points relative to the device itself. The joints which are obstructed and cannot be resolved are inferred from the posture of the person being tracked. The data were collected in frames. Each frame represents a posture of the person being tracked and it consists of twenty joints, namely: hip centre, spine, shoulder centre, head, left shoulder, left elbow, left wrist, left hand, right shoulder, right elbow, right wrist, right hand, left hip, left knee, left ankle, left foot, right hip, right knee, right ankle and right foot. The coordinates of each joint from a frame was concatenated together to represent a feature vector in higher dimension. Our research is focused on distinguishing normal from abnormal gait. The following two actions were considered in data collection and system evaluation: 1

walking left to right and vice versa

2

standing up and sitting down.

Four datasets which were 1

normal walking

2

abnormal walking

3

standing

4

sitting were collected from the Kinect device in an hour each day over a period of a month.

The datasets were divided randomly in to training datasets and test datasets for system evaluation. All feature vectors were scale to range [–1, 1] in order to avoid domination of some dimension to general performance of classifiers. Various SVM models such as C-SVC, ε-SVR, ν-SVR and SVDD were tested. However, the first three types of SVM gave similar and higher results than SVDD. So the C-SVC was chosen and the details of the C-SVC have been presented in the previous section. The pattern recognition that achieved the highest results was for the datasets consisting of height, shoulder width, and arms coordinates using SVM scaling. The Kinect device shows potential to be used in senior health monitoring due to its versatility, size and cost. As falls are a major cause of death in the elderly people, and falls account for over 80% of all injury-related admissions to hospital of people over 65 years, our proposed monitoring system will have significant contribution to health services.

358

5

W. Ma et al.

Opportunities and challenges

The ubiquitous nature of digital technologies offers great opportunities to deliver high quality personal healthcare services in real-time. However, it is inevitable that multiple source data will produce a heterogeneous overlay containing different protocols, contextual information, ad-hoc device registration, real-time data processing, and high volumes of event traffic. The data may also expose conflicting and inconsistent information. We are at the door step of exploiting the great potential of the technologies. Our work so far is only the beginning of a long journey on this road. At this stage, we can envisage and prepare to meet five types of challenges: the collection of the data and the noise levels of the data, the processing of the data, the fusion of the processed data, the validity scope of the processed information, and, most importantly, the collaboration between IT professionals and health professionals. The ubiquitous nature of the devices makes it possible to perform the data collection on a large scale, at personal level, and also in real-time life environments, of the states of the physical body and mind of an individual. However, the collection of the data is not subject to any standard protocols, and only at the mercy of the habit of the individual. For example, a smartphone, which can collect accelerometer data, could be in its owner’s pocket or handbag. The location of the pocket is not consistent and is subject to the clothes the person wearing and also the choice of the pocket at the moment. Similarly, the way the handbag is carried is not consistent either, yet it has great impact on the location of the smartphone. In addition, at any time, the owner may pick up the smartphone to make phone calls, write and read messages, or browse the internet, etc., or just simply carry the smartphone in his/her hand for a while. The location of the smartphone decides the patterns and the quality of the accelerometer data collected. This is the first great challenge we are facing: without the prior knowledge of the environments of the sensors, will we still be able to process the data collected? Another side effect of the data being collected in real-time life environments, worsened by the absence of standard protocols on the use of the devices, is the noise level of the data. The noise comes from the environments, the locations of the sensors being mounted, the disturbance to the sensors, and the sensors themselves (their quality, the models, and manufacturers, etc.). This is the second great challenge we are facing: how can we effectively filter out the noise from the signals? The comprehensive collection of the data poses another type of challenges, processing the data. On one hand, the volume of the data is massive, and also mingled with noises. On the other hand, the data are of different types and are very complicated. For example, the data collected by an accelerometer are in three dimensions and are often sampled at 100 Hz sampling rate. The EEG data collection devices range from having 32 channels to just a single channel. The signals we collected via the Emotiv EPOC headsets are of 14 channels at 128 Hz sampling rate. To make any sense of the data, the first step is to extract useful features. This step consists of further two questions: what are the useful features for the intended purposes and how to extract them. Therefore, the third great challenge we are facing is: how can we effectively extract useful features from the raw data? The massive amount of data requires computational power and also electric power to process. Unfortunately, the computational power and the electric power of these ubiquitous devices are all very limited. A distributed system thus has to be in place to

Information technologies

359

collectively achieve the goals of the proposed personal healthcare systems. In a distributed system, the communication among the involved parties is unavoidable. Regrettably, again, the communication consumes both the computational power and the electric power of the device. In addition, the communication, in terms of both technique network bandwidth and financial cost, is very expensive. The fourth challenge we are facing is: how can we distribute the tasks of the personal healthcare system in a collective environment with minimum cost? The data collected and the information obtained after processing the data are from different sensors. They may not agree with each other. Some may complement each other, some may correlate with each other, yet some may contradict with each other. How to intelligently fuse together the information from different sensors is the fifth challenge we are facing. Our last challenge is on the validity scope of the information (measurements) obtained after processing the data collected. If, as anticipated, we can obtain some objective measurements on personal well-being, the validity scope of the measurements has to be verified and established. The existing human biological measurements suggest two types of validity scopes. Some measurements are valid for the whole population, for example, blood pressure, body temperature and heart rate when the body is in idle. While the others are only valid to the individual concerned, for example facial recognition features and voice features of a person. At this stage, we are not sure of the validity scope of these measurements. For example, is the score for sway of human bodies valid to the individual or the general population? The sixth challenge we anticipate at the moment therefore is: what are the most valid measurements we can establish for the well-being of people? Finally, the greatest challenge is the collaboration between IT professionals and health professionals. Both professionals tend to work with their own comfortable zones and have limited understanding of the other domain. Bringing a team of IT professionals and health professionals together to work on the same projects requires the right people at the right time. It also takes a great amount of effort and willingness from each side to understand the other side. It took us over a year to have our team of computer scientists, health professionals, and physiotherapists to start to work together, yet we are still at the phase of understanding each other’s ways of thinking and each other’s terminologies. We are at the very early stage of understanding the application of information technology to personal healthcare systems. As our understanding goes deeper and deeper, there will surely be more and more challenges ahead. Also, in this paper, our focus is to exploit the information technology to personal healthcare systems. The feasibility is our primary focus. Therefore, the challenges from other fronts, for example, data security and privacy protection, etc., are ignored. At the moment, we are working on specific and contained research tasks so that we can thoroughly study the fundamentals. For example, when collecting the accelerometer data, we only mount the device at the back of the body trunk, along the thoracic spine at the top of the torso. We also collected accelerometer data when a person is on a treadmill to minimise the interference of other factors. After fully understanding the patterns of the accelerometer data, we can gradually introduce the other variables, for example, walking or jogging on uneven ground, the device mounted at different locations, and the sudden change of the environments, e.g., taking phone calls.

360

6

W. Ma et al.

Summary and discussion

In this paper, we have reported our preliminary work on applying technology, such as EEG devices, accelerometer and GPS devices, and Kinect, to personal healthcare. We aim to develop systems for measuring the levels of human brain activities based on EEG signals. This measurement can then help us to understand objectively, the effectiveness of meditation and also the benefits of meditation to the QOL. In addition, the measurement can also be used to detect mental fatigue and boredom and comprehend human emotion. In this study, a machine learning system is used to analyse data collected by accelerometer and GPS devices to investigate human body movement for the purposes of athlete performance enhancement, real time monitoring, estimating the levels of physical exercises, leg muscle injury prevention and recovery monitoring, and also fall prevention. The practicability of using Kinect devices to monitor the movement of aged persons to prevent falls merits further research. The information technology has made it possible for around the clock and personalised healthcare. It is up to us to work together to harness the great potential. Therefore, this work needs computer scientists, health professionals, and physiotherapists to work together to exploit the full potential of applying the technologies to health. This paper has also presented the opportunities and challenges of applying the information technology to the human mental and physical health.

References Andonegi, G. (2006) ‘HEBE: detection of falls and monitoring of the elderly’, [online] http://www.eurekalert.org/pub_releases/2006-04/ef-hdo042406.php (accessed 21 March 2012). Brophy, S., Hill, R., Zhou, S.M. and Lyons, R. (2012) ‘MIA: measuring infant activity – the pilot study’, MRC Population Health Methods and Challenges Conference, Birmingham, UK. Davidson, R., Kabat-Zinn, J., Schumacher, J., Rosenkranz, M., Muller, D., Santorelli, S., Urbanowski, F., Harrington, A., Bouns, K. and Sheridan, J. (2003) ‘Alterations in brain and immune function produced by mindfulness meditation’, Psychosomatic Medicine, Vol. 65, No. 4, pp.564–570. Dressler, O., Schneider, G., Stockmanns, G. and Kochs, E.F. (2004) ‘Awareness and the EEG power spectrum: analysis of frequencies’, British Journal of Anaesthesia, Vol. 93, No. 6, pp.806–809. Emotiv (2012) [online] http://www.emotiv.com/ (accessed 17 March 2012). Freeman, W.J. (2000) ‘A neurobiological interpretation of semiotics: meaning, representation, and information’, Information Sciences, Vol. 124, No. 1, pp.93–102. GPSports (2012) ‘SPI PRO X II’ [online] http://gpsports.com/gpsnew/spiprox.php (accessed 15 March 2012). Hamadicharef, B., Zhang, H.H., Guan, C.T., Wang, C.C., Phua, K.S., Tee, K.P. and Ang, K. K. (2009) ‘Learning EEG-based spectral-spatial patterns for attention level measurement’, 2009 IEEE International Symposium on Circuits and Systems (ISCAS2009), Taipei, Vol. 6, pp.806–809. Hölzel BK., Vangel, M., Congleton, C., Yerramsetti, SM., Gard, T. and Lazar, SW. (2011) ‘Mindfulness practice leads to increases in regional brain gray matter density’, Psychiatry Research, Vol. 191, No. 1, pp.36–43. Kamen, G., Patten, C., Du, C.D. and Sison, S. (1998) ‘An accelerometry-based system for the assessment of balance and postural sway’, Gerontology, Vol. 44, No. 1, pp.40–45.

Information technologies

361

Kannus, P., Khan, K.M. and Lord, S.R. (2006) ‘Preventing falls among elderly people in the hospital environment’, The Medical Journal of Australia, Vol. 184, No. 8, pp.379–382. Kinect (2011) ‘Kinect for windows SDK beta’ [online] http://research.microsoft.com/enus/collaboration/focus/nui/kinect-windows.aspx (accessed 21 March 2012). Lane, N.D., Miluzzo, E., Lu, H., Peebles, D., Choudhury, T. and Campbell, T.A. (2010) ‘A survey of mobile phone sensing’, IEEE Communications Magazine, Vol. 48, No. 9, pp.140–150. Lee, R.Y.W. and Carlisle, A.J. (2011) ‘Detection of falls using accelerometers and mobile phone technology’, Age & Ageing, Vol. 40, No. 6, pp.690–696. Lin, H. (2010) ‘Measurable meditation’, Proceedings of the International Symposium on Science 2.0 and Expansion of Science (S2ES 2010), Orlando, Florida, USA, pp.56–61. Lin, Y-P., Wang, C-H., Jung, T-P., Wu, T-L., Jeng, S-K., Duann, J-R. and Chen, J-H. (2010) ‘EEG-based emotion recognition in music listening’, IEEE Transactions on Biomedical Engineering, Vol. 57, No, 7, pp.1798–1806. NeuroSky (2012) [online] http://www.neurosky.com/ (accessed 17 March 2012). O’Sullivan, M., Blake, C., Cunningham, C., Boyle, G. and Finucane, C. (2009) ‘Correlation of accelerometry with clinical balance tests in older fallers and non-fallers’, Age & Ageing, Vol. 38, No. 3, pp.308–313. Oh, B., Butow, P., Mullan, B. and Clarke, S. (2008) ‘Medical qigong for cancer patients: pilot study of impact on quality of life, side effects of treatment and inflammation’, American Journal of Chinese Medicine, Vol. 36, No. 3, pp.459–472. Oh, B., Butow, P., Mullan, B., Clarke, S., Beale, P., Pavlaki, s.N., Lee, M., Rosenthal, D., Larkey, L. and Vardy, J. (2012) ‘Effect of medical qigong on cognitive function, quality of life, and a biomarker of inflammation in cancer patients: a randomized controlled trial’, Journal of Supportive Cancer Care, Vol. 20, No. 6, pp.1235–1242. Oh, B., Butow, P., Mullan, B., Clarke, S., Beale, P., Pavlakis, N., Kothe, E., Lam, L. and Rosentha, l.D. (2009) ‘Impact of medical qigong on quality of life, fatigue, mood and inflammation in cancer patients: a randomized controlled trial’, Ann Oncol, Vol. 21, No. 3, pp.608–614. Oh, B., Butow, P., Mullan, B., Clarke, S., Beale, P., Pavlakis, N., Kothe, E., Lam, L. and Rosenthal, D. (2010) ‘Impact of medical qigong on quality of life, fatigue, mood and inflammation in cancer patients: a randomized controlled trial’, Annals of Oncology, Vol. 21, No. 3, pp.608–614. Records, G.W. (2011) ‘Fastest-selling gaming peripheral’, retrieved 14 December 2011 [online] http://www.guinnessworldrecords.com/records-9000/fastest-selling-gaming-peripheral/. Ross, S. and Guskiewicz, K.M. (2003) ‘Time to stabilization: a method for analyzing dynamic postural stability’, International Journal of Athletic Therapy & Training, Vol. 8, No. 3, pp.37–39. Ross, S.E., Guskiewicz, K.M. and Yu, B. (2005) ‘Single-leg jump-landing stabilization times in subjects with functionally unstable ankles’, Journal of Athletic Training, Vol. 40, No. 4, pp.298–304. Simkevitz, H. (2009) ‘Why privacy matters in health care delivery: a value proposition’, The Proceedings of CONGRESS '09, World Congress on Privacy, Security, Trust and the Management of e-Business, 2009, 25–27 August 2009, Saint John, NB, Canada, pp.193–201. TM ‘Transcendental Meditation’ [online] http://www.britannica.com/EBchecked/topic/602436/ Transcendental-Meditation (accessed 21 March 2012). Waarsing, J.H., Mayagoitia, R.E. and Veltink, P.H. (1996) ‘Quantifying the stability of walking using accelerometers’, 18th Annual International Conference of the IEEE on Bridging Disciplines for Biomedicine, pp.469–470. World Health Organization (WHO) (2012) Global Strategy on Diet, Physical Activity and Health [online] http://www.who.int/dietphysicalactivity/factsheet_inactivity/en/index.html (accessed 29 May 2012).

362

W. Ma et al.

Yoshida, T., Mizuno, F., Hayasaka, T., Tsubota, K., Wada, S. and Yamaguchi, T. (2006) ‘Gait analysis for detecting a leg accident with an accelerometer’, First Transdisciplinary Conference on Distributed Diagnosis and Home Healthcare (D2H2), pp.43–46. Zhang, S., McCullagh, P., Nugent, C. and Zheng, H. (2010) ‘Activity monitoring using a smart phone’s accelerometer with hierarchical classification’, Sixth International Conference on Intelligent Environments. Zhou, S-M., Gan, J.Q. and Sepulveda, F. (2008) ‘Classifying mental tasks based on features of higher-order statistics from EEG signals in brain-computer interface’, Information Sciences, Vol. 178, No. 6, pp.1629–1640.