Intelligent Health Monitoring Based on Pervasive ...

May 5, 2014 16:50 WSPC/INSTRUCTION FILE

S021821301460001X

International Journal on Artificial Intelligence Tools Vol. 23, No. 3 (2014) 1460001 (17 pages) c World Scientific Publishing Company

DOI: 10.1142/S021821301460001X

Intelligent Health Monitoring Based on Pervasive Technologies and Cloud Computing

Ilias Maglogiannis∗ Int. J. Artif. Intell. Tools 2014.23. Downloaded from www.worldscientific.com by Prof. Ilias Maglogiannis on 06/10/14. For personal use only.

Department of Digital Systems, University of Piraeus Grigoriou Lampraki 12618532 Piraeus, Greece [email protected]

Charalampos Doukas Department of Information and Communication Systems Engineering University of the Aegean, Karlovasi Samos GR-83200, Greece [email protected]

Received 21 April 2013 Accepted 6 December 2013 Published 28 May 2014

The proper management of patient data and their accessibility are still remaining issues that prevent the full deployment and usage of pervasive healthcare applications. This paper presents an integrated health monitoring system based on mobile pervasive technologies. The system utilizes Cloud Computing for providing robust and scalable resources for sensor data acquisition, management and communication with external applications like health information systems. A prototype has been developed using both mobile and wearable sensors for demonstrating the usability of the proposed platform. Initial results regarding the performance of the system, the efficiency in data management and user acceptability have been quite promising. Keywords: Patient monitoring; wearable sensors; sensor data management; motion classification; cloud computing.

1.

Introduction

The proper delivery of healthcare services among with patient monitoring are considered key issues for improving the quality of life and ensuring efficient health and social care. Mobile pervasive healthcare technologies can support a wide range of applications and services, including mobile telemedicine, patient monitoring, location-based medical services, emergency response and management, personalized monitoring and pervasive access to healthcare information, providing great benefits to both patients and medical personnel.1,2 The realization, however, of health information management through mobile devices introduces several challenges, like data storage and management (e.g., physical storage issues, availability and maintenance), interoperability and availability of ∗ Corresponding

author 1460001-1

Int. J. Artif. Intell. Tools 2014.23. Downloaded from www.worldscientific.com by Prof. Ilias Maglogiannis on 06/10/14. For personal use only.

I. Maglogiannis & C. Doukas

heterogeneous resources, security and privacy (e.g., permission control, data anonymity, etc.), unified and ubiquitous access. One potential solution for addressing all aforementioned issues is the introduction of Cloud Computing concept in electronic healthcare systems. Cloud Computing provides the facility to access shared resources and common infrastructure in a ubiquitous and pervasive manner, offering services ondemand, over the network, to perform operations that meet changing needs in electronic healthcare application. In this context we have developed an integrated system for health and patient context monitoring based on mobile and wearable sensors that utilizes a Cloud computing infrastructure for data management. The rest of the paper is structured as follows: Section 2 presents related work and discusses background information about Cloud computing and healthcare. Section 3 presents the proposed system architecture and the corresponding software and hardware modules. Section 4 describes an initial implementation of a prototype system and Section 5 presents evaluation results in terms of performance and user acceptance. Finally, Section 6 concludes the paper. 2.

Related Work and Background Information

The application of mobile devices for pervasive healthcare information management has already been acknowledged and well established.1,8 Authors in 3 present the benefits of using virtual health records for mobile care of elderly citizens. The main purpose of the work is to provide seamless and consistent communication flow between home health care and primary care providers using devices like PDAs and Tablet PCs. Smart cards and web interfaces have been used in 4 for storing patient records electronically. The MADIP system5 is a distributed information platform allowing widearea health information exchange based on mobile agents. In Ref. 2 authors present a mobile platform for exchanging medical images and patient records over wireless networks using advanced compression schemes. The majority of the aforementioned works is based on proprietary architectures and communication schemes and requires the deployment of specific software components. Furthermore, these works focus mostly on delivering data to healthcare applications and do not address issues of data management and interoperability issues introduced by the heterogeneous data resources found in modern healthcare systems. Cloud Computing is a model for enabling convenient, on-demand network access to a shared group of configurable computing resources (e.g., networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management effort or service provider interaction. Resources are available over the network and accessed through standard mechanisms that promote use by heterogeneous thin or thick client platforms (e.g., smart phones). Examples of resources include storage, processing, memory, network bandwidth, and virtual machines. Given the characteristics of Cloud Computing and the flexibility of the services that can be developed, a major benefit is the agility that improves with users being able to rapidly and inexpensively reprovision technological infrastructure resources. Device and location independence 1460001-2


Pervasive Health Monitoring and Cloud Computing

enable users to access systems using a web browser, regardless of their location or what device they are using (e.g., mobile phones). Multi-tenancy enables sharing of resources and costs across a large pool of users, thus allowing for centralization of infrastructure in locations with lower costs. Reliability improves through the use of multiple redundant sites, which makes Cloud Computing suitable for business continuity and disaster recovery. Security typically can be improved, due to centralization of data and increased availability of security-focused resources. Sustainability comes about through improved resource utilization, resulting in more efficient systems. The usage of Cloud Computing provides data management and access functionality overcoming the data management and access issues as discussed in previous sections. The concept of utilizing Cloud Computing in the context of healthcare information management is relatively new but is considered to have great potential.9 A number of Cloud-based services dedicated for storing sensor-based data are nowadays available. Pachube,15 Nimbits,18 ThingSpeak16 and iDigi17 are a few that could be mentioned. These services however, focus mostly on the visualization of the data and usually lack of secure data access and provision of interfaces for linkage to mobile or external applications for further processing. Neither of the referred services is designed for hosting patient data. Finally, the acquisition of healthcare data (like heartbeat, body temperature, oxygen saturation and related biosignal) and patient context (like location, outside temperature, activity status, etc.) can be performed through various body sensors and mobile technologies. Table 1 presents an overview of the most commonly used biosignal in biomedical application and important information about their acquisition.22,23 All the aforementioned biosignal can be directly interfaced with the analog input ports of various microcontroller platforms. Depending on the number of sensors used and the type of the biosignal the patient status can be indicated by transmitting the signals to a monitoring unit where medical experts can decide about the condition of the patient. The usage of alternative sensors (like accelerometers and gyroscopes) for the determination of the physical status of the patient has already been proposed in the literature.12,21 The proposed architecture is open and interoperable with respect to the sensors that can be integrated into the system. The sensor gateways are based on open microcontroller Table 1.

Common biosignal and important information.

Biomedical Measurements (Broadly Used Biosignal)

Voltage Range (V)

Number of Sensors

Information Rate (kb/s)

ECG Heart sound Heart rate EEG EMG Photoplethysmography (SpO2) Respiratory rate Temperature of body

0.5–4 m Extremely small 0.5–4 m 2–200 µ 0.1–5 m Small Small 0–100 m

5–9 2–4 2 20 2+ 2 1 1+

15 120 0.6 4.2 600 1.4 0.8 0.08

1460001-3


platforms (like the Arduino11) that provide both analog and digital interfaces for communicating with various types of sensors, even implantable devices like the pulseoxymeter presented in 24). 3.

Materials and Methods — The Proposed System Architecture and the Corresponding Modules


This section presents the proposed architecture for developing and deploying the integrated monitoring system. It also describes the acquisition of healthcare data using body sensors and mobile technologies. It discusses networking issues between the sensors and the Cloud infrastructure and also provides information about the pattern recognition and patient context monitoring. 3.1.

Overall architecture

This proposed architecture for acquiring and managing healthcare data on the Cloud using mobile pervasive technologies is illustrated in the Figure 1. The main components of the architecture are: • •

•

•

•

The wearable and mobile sensors that acquire patient biosignals, motion and contextual information. The sensor gateway that collects all the signals from the sensors and forwards them to the Internet. It can be a mobile phone or a microcontroller platform capable of communicating with the Internet. It also forwards information about the status of the sensors (e.g., proper operation, power source levels, etc.). The sensor gateway features in addition temporal data storage so that crucial sensor information can be retransmitted in case of connection signal problems or other communication issues. The communication APIs that are provided by the Cloud platform. The latter are lightweight interfaces (like REST Web Services) than can be used by the sensor gateways for sending sensor data and retrieving information. The API can also be utilized by external applications for data processing, alert management, billing, etc. The managing application consists of a web-based application that is updated real time and provides visualization of the sensor data (in graphs, etc.) and important information about the patient’s context (like location, activity status, etc.). The Cloud infrastructure that hosts the interfaces and the managing application. It provides the essential resources (like CPU, storage and application servers) for deploying the web application and the interfaces that enable the communication with the sensors and the various external systems.

Every communication that takes place between the Cloud infrastructure and the rest of the components is secure by applying appropriate authentication and data encryption mechanisms. Sensors can be authenticated by unique id and data can be encrypted using symmetric encryption techniques.13 Users and external applications can be authenticated using more sophisticated mechanisms like PKI and digital signatures.14

1460001-4



Fig. 1. The proposed architecture: Appropriate APIs and web-based applications deployed on the Cloud provide the essential communication channels for sensors, caregivers and business stakeholders.

The major features of the proposed architecture are its scalability, interoperability and lightweight access. It is scalable due to the fact that it relies on a Cloud infrastructure that provides resources based on utilization and demand. More users, sensors and other data sources can be added without affecting the functionality of the system or without the need for further maintenance or expansion. The web-services based interfaces ensure the maximum interoperability with external applications. The REST API is very lightweight and can be easily accessed and implemented by wireless sensor and mobile platforms.25.26 3.2.

Networking and data management issues

The transmission of the biosignal data to the Cloud requires a wireless networking infrastructure with Internet access. Table 2 presents an overview of the most common wireless transmission technologies. Considering the bandwidth requirements of biosignal

1460001-5

I. Maglogiannis & C. Doukas Table 2.

Common wireless transmission technologies.


Technology

Data Rate

Range

IEEE 802.11a

54 Mbps

150 m

IEEE 802.11b

11 Mbps

150 m

Bluetooth (IEEE 802.15.1) HiperLAN2 HomeRF (Shared Wireless Access Protocol, SWAP)

721 Kbps 54 Mbps 1.6 Mbps (10 Mbps for Ver.2)

10 m–150 m 150 m 50 m

DECT IEEE 802.15.3 (high data rate wireless personal area network) IEEE 802.15.4 (low data rate wireless personal area network), Zigbee

32 kbps 11–55 Mbps

100 m 1 m–50 m

250 kbps, 20 kbps, 40 kbps

100 m–300 m

Cellular 3G

14.4 Mbps/ 5.8 Mbps

m–km

data (see Table 1), all existing technologies can provide the required data rates. However data range and power consumption are the most important issues that determine the adoption of a specific technology. For instance, implantable and textile sensor devices are usually equipped with low power communication technologies like 802.15.427 or custom Radio Frequency (RF) solutions like in Ref. 2828. The latter technologies however do not provide direct connectivity to the Internet. Thus the sensor gateway can be provided with the essential interfaces and a WiFi or GSM Internet connection for providing the essential access to the Cloud infrastructure. The latter module can also provide the essential security for the communication with the Internet. Signal data can be encrypted using symmetric encryption techniques and transferred over secure communication channels like the SSL or extensions to the latter specially designed for medical data transmission.29 Regarding authentication and user access on the data, various schemes can be implemented on the Cloud applications that conform to related standards (like HIPAA).30,31 Regarding data management, the patient signals acquired from sensors have both storage and availability requirements. The sensors generate data with different information rates (see Table 1) that need to be assessed by medical personnel in various locations (e.g., treatment centers, being mobile, etc.). In addition, according to the significance of the biosignal, the transmission might have to be instant and in real time or reactively (i.e. when a sensor values exceeds a threshold like when body temperature exceeds 37oC). All data must be instantly accessible to users and stored also for further processing (e.g., automated assessment of context and status) and for future reference (e.g., as part of patient’s medical record). The Cloud infrastructure provides both the accessibility (through hosting web applications and communication interfaces) and the essential computational and storage resources. The data are stored into a (relational or not) database for better maintenance and searchability. Due to the offered scalability, the platform administrators do not have to modify the infrastructure for hosting more sensors, data or users. Upon request the resources are automatically provided to the system. The 1460001-6


Cloud platform provides also the essential load balancing, backup and restore mechanisms. System administrators need only to focus on functional maintenance of the system like the interfacing with the sensors or the external applications. It should be noted that the interoperability feature offered by Cloud platforms, enables the usage of any terminal device (i.e smart phone, tablet, laptop) independently of the installed operating system.


3.3.

Introducing intelligence — Monitoring of patient context

Pattern recognition and data classification have already been established techniques in patient context awareness.32,33 There is a vast array of established classification techniques, ranging from classical statistical methods, such as linear and logistic regression, to neural network and tree-based techniques (e.g., feed-forward networks, which includes multilayer perception, Radial-Basis Function networks, Self-Organizing Map, or Kohonen-Networks), to the more recent Support Vector Machines. Other types of hybrid intelligent systems are neuro-fuzzy adaptive systems, which can comprise of an adaptive fuzzy controller and a network-based predictor. More information regarding data classification techniques can be found in Ref. 36. Context awareness refers to the capability of the computing or networking applications to be aware of the existence and characteristics of the user's activities and environments. In rapidly changing scenarios, such as the ones considered in the fields of mobile, pervasive, or ubiquitous computing, systems have to adapt their behavior based on the current conditions and the dynamicity of the environment they are immersed in. A system is context-aware if it can extract, interpret and use context information and adapt its functionality to the current context of use. The challenge for such systems lies in the complexity of capturing, representing and processing contextual data. The way contextaware applications make use of context can be categorized into the three following classes: presenting information and services, executing a service, and tagging captured data. Presenting information and services refers to applications that either present context information to the user, or use context to propose appropriate selections of actions to the user. Automatically executing a service describes applications that trigger a command, or reconfigure the system on behalf of the user according to context changes. Attaching context information for later retrieval refers to applications that tag captured data with relevant context information. More information on context aware systems can be found in Refs. 40 and 41. Medical data analysis has also moved towards the Cloud for more efficient data management.34 The Cloud can host applications based on Java EE and libraries that provide data classification methods, like the core of the Weka classification system.35 Thus activity recognition and context awareness applications like in Refs. 12, 37–41 can be hosted by the proposed solution. As discussed in the next Section the proposed system may use one or several AI modules to classify specific activities, their intensity (i.e high, medium, low) and detect emergency situations (i.e patient falls). Appropriate training and calibration of the AI modules is required in advance. 1460001-7


4.

Architecture Prototype

To demonstrate the functionality and main features of the proposed architecture we have developed a system that consists of wearable and mobile sensor monitoring and data management through a Cloud infrastructure. The system consists of three parts: A wearable/textile part that has been designed for maximum user convenient, a mobile part that adds additional patient context awareness and the Cloud infrastructure that provides all the essential interfaces for collecting the data and tools for managing the latter.


4.1.

The wearable part

For the wearable part we have used textile accelerometers and a heartbeat chest strap by Polar.10 The latter sensors are connected to a textile version of the Arduino open hardware microcontroller platform, called LilyPad.11 Arduino is an open-source singleboard microcontroller. The hardware consists of a simple open hardware design for the Arduino board with an Atmel AVR processor and on-board I/O support. The software consists of a standard programming language compiler and the boot loader that runs on the board. The LilyPad Arduino is a microcontroller board designed for wearables and e-textiles. It can be sewn to fabric and similarly mounted power supplies, sensors and actuators with conductive thread (see Figure 2). For the connection between the Polar monitor and the microcontroller, the Polar HeartRate Module has been utilized. Lilypad collects data through the appropriate embedded software and transmits them on an Android-based mobile phone through a Bluetooth interface. An appropriate application has developed for the Android that collects the data and forwards them to the Cloud (see Figure 3). The application stores also the acquired data on the mobile’s memory card for future usage of retransmission in case of communication problems. All textile sensors and the microcontroller have been sewed on a sock that can be worn easily by the user. The first prototype has been presented in Ref. 19.

Fig. 2.

The CloudSensorSock and the main hardware modules as sewed on the final prototype. 1460001-8


For the initial implementation a different Cloud infrastructure than the presented one (see sub-section C) has been utilized. The mobile application has communicated seamlessly with the new infrastructure providing the same functionality and performance.


4.2.

The mobile part

The mobile part has also been developed using the Arduino platform. It consists of additional sensors like an air quality sensor and wireless networking modules that provide direct connectivity to the Internet. The modules can be either a WiFi or a 3G/GPRS modem depending on whether indoor and/or outdoor functionality is provided. The purpose of developing and using an additional mobile sensor unit is twofold: (a) to demonstrate the feasibility of existing microcontroller platforms in communicating directly with the Internet and the presented system and (b) more sensors can be added ad-hoc on the system since both the hardware (the Arduino platform) and the developed software provide such features. The mobile part has been initially presented and evaluated as an individual monitoring solution in Ref. 20. Due to the interoperability of the platform there was no additional effort needed for integrating the application into the new Cloud infrastructure. The latter proves also the modularity of the system; new features and applications can be connected to the system without any special requirements for system re-design or re-deployment. A screenshot of the mobile application is illustrated in Figure 4.

Fig. 3.

4.3.

The mobile part consisting of an arduino board, motion sensors and the air quality sensors.

The Cloud infrastructure

For the Cloud infrastructure we have initially utilized the Google App Engine, as an open source way for hosting online applications on the Cloud. It is a cloud-computing platform for developing and hosting web applications in Google-managed data centers. Google App Engine makes it easy to build an application that runs reliably, even under heavy load and with large amounts of data. The main features of the Google App Engine include dynamic web serving, with full support for common web technologies, automatic scaling and load balancing of the applications and APIs for authenticating users and sending email using Google Accounts. 1460001-9



Fig. 4. The mobile application interface demonstrating the heart beat rate and history graph based on measurements by the mobile part.

Sensor data are retrieved using a lightweight REST API that enables direct communication from the Arduino Lilypad microcontroller when direct wireless connectivity is available. The data are stored using the App Engine’s Datastore feature. The Datastore writes data in objects known as entities, and each entity has a key that identifies the entity. Entities can belong to the same entity group, which allows you to perform a single transaction with multiple entities. Entity groups have a parent key that identifies the entire entity group. App Engine's infrastructure takes care of all of the distribution, replication and load balancing of data. Due to some limitations of the Google App Engine (e.g., limited ability for creating local files and absence of database schemas) the application has been ported to the Jelastic Cloud service. The Jelastic44 is a Platform as a Service (PaaS) Cloud provider that allows users to deploy Java-based applications providing all the essential components (application server instances, databases, load balancers, etc.) and all the appropriate scalability. Jelastic provides full access to the application server runtime environment, which enables the deployment of additional Java extensions like encryption and authentication libraries. For the specific application the Tomcat application server among with a MySQL database have been utilized. Data encryption has been achieved using the java cryptographic extension implementing a symmetric (AES) data encryption. AES has been also implemented on the Arduino so that sensor data are encrypted upon acquisition. The Weka core library has also incorporated into the system allowing the deployment of the motion analysis platform and context awareness platform described in Refs. 12 and 37.

1460001-10



Fig. 5.

The graphical interface of the web management application on the Cloud.

The system is able to acquire motion, heartbeat and air quality data and display them in real time on the appropriate graphical interface (see Figure 5). It also performs real time classification of the motion data and provides an estimation of the user’s activity and alerts in case of an emergency indication (e.g., a fall incident estimation). 5.

Initial Evaluation Results

In the context of the initial system evaluation 5 users have been equipped with the sensors described in Section 4. All of them (average males aged between 25–35) have used both the wearable and mobile monitoring units while performing every day activities. The sampling rate for the accelerometer values has been at 10 samples per second (indicated as an acceptable rate for identified falls by previous works as in Ref. 12, for heartbeat data once per heartbeat detected and for air quality data every 1 minute. The mobile gateway in each case is used to transmit the data over a commercial 3G network to the cloud infrastructure. The differences between two sequential acceleration values have been transmitted at the same rate whereas the average heartbeats per minute have been calculated and transmitted every second respectively. During the initial experimentation with the system, a drop packet rate of 20–30% has been detected. This fact is either due to the Arduino low resources for high rate sampling of sensors and transmitting the data at the same time, or due to network congestion because of the repetitive REST calls at such a high sampling rate (i.e. 10 acceleration samples per second). In order to address this issue, a memory buffer has been 1460001-11



introduced on the Arduino side that collects motion data during a 10 second time frame and then transmits the latter to the Cloud. This way the drop rate has been minimized between 2–5%, which is quite acceptable for the application. Regarding power consumption, the wearable part (powered by one AAA battery) was able to last for at least 18 hours wile the mobile part could not last more than 6. This difference can be explained by the more energy-intensive communication mechanism (3G) of the mobile unit compared to the low-power Bluetooth, and the power consumption of the air quality sensor. The system has been quite successful in identifying the user’s activities. More specifically, activities based on motion analysis have been classified into three categories: high activity (like running), medium activity (like walking) and low activity (e.g., lying on bed or sitting). Initially a train model has been built using the Weka core library and modifying the Android mobile application for allowing users to annotate first the type of performed activity. All 5 users have been asked to use the sensors to collect data performing all three types of motion: the train model has been built on 20 minutes of jogging and fast walking, 30 minutes of walking and 60 minutes of performing low activity. The mobile application has transmitted the data and the corresponding motion type as selected by the user. After collecting the data the training model has been built on the Cloud. A number of different classifies have been evaluated using the initial motion and heartbeat data. The features utilized for building the classification models are the acceleration changes in the X, Y and Z axis, the input from the gyroscope (tilt) sensors (a tilt value for each axis) and the average heartbeat rate (beats per minute). Four classes were examined, namely jogging, fast walking, walking and sitting/low activity). To evaluate each classifier, the original data set has been divided into a training set containing the 66% of the dataset and a test set containing the rest 34% of the dataset. The evaluation results are presented in the following Table 3. The feature of classification training on the Cloud provides a great benefit for the expandability and efficiency of the system: when new users are added, a custom train model can be developed on demand maximizing performance and accuracy. A time frame of ten (10) second is used for performing the classification task and updating the detected motion type, while all the data are time stamped. Table 3.

Performance evaluation of motion classification.

Classifier

Accuracy (%)

ROC Area

BayesNet

79.72

0.937

Logistic Regression47

70.27

0.814

Support

Vector48

66.21

0.822

KStar 49

82.43

0.956

Classification via Regression50

82.43

0.941

83.78

0.931

86.48

0.938

PART decision NBTree46

list45

1460001-12

Pervasive Health Monitoring and Cloud Computing Table 4.

MOS for the usability, convenience and functionality of the prototype system.

Impact Category


Usability Performance Convenience

Interface

Mobile Unit

Wearable Unit

4.3 4.5 4.0

4.0 4.3 4.5

4.5 4.3 3.0

Based on the evaluation results (accuracy in motion type characterization up to 86.48%) the system can be considered quite successful in providing an estimation of the user’s activity and context. In addition, we have requested from potential users of the system (both patient that need remote supervision and physicians) to assess the platform in terms of usability, user convenience and functionality. Table 4 presents the Mean Opinion Score (MOS) results (in the range of 1 to 5) for the usability, convenience and functionality for the user interface, the mobile unit and the wearable unit respectively. The user acceptance regarding the usability and performance of system modules appears to be quite positive. The wearable part is considered to be the less convenient (especially compared to the mobile unit) because according to users, the mobile device appeared more concrete than the wearable prototype. The greater usability impact however shows the potential of wearable devices in pervasive healthcare monitoring. 6.

Discussion and Conclusion

The costs of health care impose an enormous burden on the economy. The latest projections from the Centers for Medicare & Medicaid Services show that annual healthcare expenditures are expected to reach $3.1 trillion by 2012, growing at an average annual rate of 7.3% during the forecast period or 17.7% of gross domestic product, up from 14.1% today.42 Recent advances in communication and information technologies have impulse the development of novel tools that enable remote management and monitoring of chronic disease, emergency conditions, and the delivery of health care on patient’s site, saving time, travel and other expenses. Health monitoring in home environments can be accomplished by establishing ambulatory monitors that utilize wearable sensors and devices that record physiological signals, sensors embedded in the home environment to collect behavioral and physiological data or a combination of the latter. Studies for such non-invasive monitoring technologies have shown good acceptance rates by patients, presenting overall a positive impact on their perceived quality of life, as well as reducing hospitalization costs.43 This need however for pervasive healthcare applications has also introduced issues like proper management of the sensor data, analysis and processing. Cloud computing through its elasticity and facility to access shared resources and common infrastructure in a ubiquitous and pervasive manner is a promising solution for efficient management of pervasive healthcare data. The presented system is unique as a dedicated solution for managing patient-related data on the cloud and that utilizes both open hardware and open software resources for 1460001-13



developing the hardware and software parts of the platform. It allows direct communication of the sensor devices with the Cloud application due to the lightweight API used, while it is highly scalable in the context of data stored, users and sensors supported. It is also highly interoperable and modular since it can integrate new modules (like the monitoring units described in the use case scenario) without any additional effort. An important aspect of the proposed system that needs to be further examined is the creation of generic classification rules for various motion types using the train set obtained by various users. Such a feature would increase the acceptance of the proposed system by a higher number of users, since it would not require the initial training phase. Key issues arising also from such a system include social issues about acceptance and training with the technology. Other open issues that need to be addressed are the security of privacy of data and the energy efficiency of the textile sensors and microcontroller platform, in order to extend the system autonomy. Acknowledgment The authors would like to thank the European Union (European Social Fund ESF) and Greek national funds through the Operational Program “Education and Lifelong Learning” of the National Strategic Reference Framework (NSRF) — Research Funding Program: \Thalis\I nterdisciplinary Research in Affective Computing for Biological Activity Recognition in Assistive Environments for financially supporting this work. References 1. Upkar Varshney, Pervasive healthcare, IEEE Computer Magazine 36(12) (2003) 138–140. 2. Maglogiannis, I., Doukas, C., Kormentzas, G., and Pliakas, T., Wavelet-based compression with ROI coding support for mobile access to DICOM images over heterogeneous radio networks, IEEE Transactions on Information Technology in Biomedicine 13(4) (2009) 458– 466. 3. Sabine Koch, Maria Hägglund, Isabella Scandurra, and Dennis Moström: Towards a virtual health record for mobile home care of elderly citizens, presented in MEDINFO 2004 (Amsterdam, 2004). 4. Alvin T. S., Chan, WWW_smart card: Towards a mobile health care management system, International Journal of Medical Informatics 57 (2000) 127–137. 5. Chuan Jun Su, Mobile multi-agent based, distributed information platform (MADIP) for widearea e-health monitoring, Computers in Industry 59 (2008) 55–68. 6. Khawar Hameed, The application of mobile computing and technology to health care services, Telematics and Informatics 20 (2003) 99–106. 7. Eneida A. Mendonça, Elizabeth S. Chen, Peter D. Stetson, Lawrence K. McKnight, Jianbo Lei, and James J. Cimino, Approach to mobile information and communication for health care, International Journal of Medical Informatics 73 (2004) 631–638. 8. Eneida A. Mendonça, Elizabeth S. Chen, Peter D. Stetson, Lawrence K. McKnight, Jianbo Lei, and James J. Cimino, Approach to mobile information and communication for health care, International Journal of Medical Informatics 73 (2004) 631–638. 9. Ofer Shimrat, Cloud Computing and Healthcare, San Diego Physician (2009), pp. 26–29. 10. Polar, www.polar.com 1460001-14



11. The Arduino LilyPad, http://arduino.cc/en/Main/ArduinoBoardLilyPad. 12. Charalampos Doukas and Ilias Maglogiannis, Emergency fall incidents detection in assisted living environments utilizing motion, sound and visual perceptual components, in IEEE Transactions on Information Technology in Biomedicine 15(2) (March 2011) 277–289. 13. Liu Zhenglin, Zeng Yonghong, Zou Xuecheng, Han Yu, and Chen Yicheng, A high-security and low-power aes s-box full-custom design for wireless sensor network, 2007 Int. Conf. on Wireless Communications, Networking and Mobile Computing (WiCom 2007) (21–25 September, 2007), pp. 2499–2502. 14. Il Kon Kim, Pervez, Z., Khattak, A. M., and Sungyoung Lee, Chord based identity management for e-healthcare cloud applications, applications and the internet (SAINT), 2010 Int. Symp. on 10th IEEE/IPSJ (19–23 July, 2010), pp. 391–394. 15. The Pachube Feed Cloud Service, http://www.pachube.com 16. Internet of Things – ThingSpeak service, http://www.thingspeak.com 17. iDigi Device Cloud, http://www.idigi.com 18. Nimbits Data Logging Cloud Sever, http://www.nimbits. 19. Charalampos Doukas and Ilias Maglogiannis, The CloudSensorSock: Managing wearable sensor data on the cloud through open source tools, presented at the IEEE EMBC 2011 Unconference Event (Boston, USA, 30 August). 20. Charalampos Doukas, Thomas Pliakas, Panayiotis Tsanakas, and Ilias Maglogiannis, Distributed management of pervasive healthcare data through cloud computing, presented at 2nd Int. IEEE — ICST Conference on Wireless Mobile Communication and Healthcare — MobiHealth 2011 (Kos island, Greece, October 2011). 21. Bouten, C. V. C., Koekkoek, K. T. M., Verduin, M., Kodde, R., and Janssen, J. D., A triaxial accelerometer and portable data processing unit for the assessment of daily physical activity, IEEE Transactions on Biomedical Engineering 44(3) (1997) 136–147. 22. Paksuniemi, M., Sorvoja, H., Alasaarela, E., and Myllyla, R., Wireless sensor and data transmission needs and technologies for patient monitoring in the operating room and intensive care unit, in Proc. of 27th Annual Int. Conf. of the Engineering in Medicine and Biology Society, 2005 (IEEE-EMBS 2005) (17–18 January, 2006), pp. 5182–5185. 23. Charalampos Doukas, Vangelis Metsis, Erick Becker, Zhengyi Le, Fillia Makedon, and Ilias Maglogiannis, Digital cities of the future: Extending @home assistive technologies for the elderly and the disabled, Telemat. Informat. (2010), doi:10.1016/j.tele.2010.08.001. 24. Reichelt, Stephan Fiala, Jens Werber, Armin Forster, Katharina Heilmann, Claudia Klemm, and Rolf Zappe, Hans, Development of an implantable pulse oximeter, IEEE Transactions on Biomedical Engineering 55(2) (2008) 581. 25. Gupta, V., Poursohi, A., and Udupi, P., Sensor.Network: An open data exchange for the web of things, 2010 8th IEEE Int. Conf. on Pervasive Computing and Communications Workshops (PERCOM Workshops) (March 29, 2010–April 2, 2010), pp. 753–755. 26. Cardei, M., Marcus, A., Cardei, I., and Tavtilov, T., Web-based heterogeneous WSN integration using pervasive communication, 2011 IEEE 30th Int. Performance Computing and Communications Conference (IPCCC) (November 2011), pp. 1–6, 17–19. 27. Valdastri, P., Susilo, E., Forster, T., Strohhofer, C., Menciassi, A., and Dario, P., Wireless implantable electronic platform for chronic fluorescent-based biosensors, IEEE Transactions on Biomedical Engineering 58(6) (June 2011) 1846–1854. 28. Majerus, S. J. A., Fletter, P. C., Damaser, M. S., and Garverick, S. L., Low-power wireless micromanometer system for acute and chronic bladder-pressure monitoring, IEEE Transactions on Biomedical Engineering 58(3) (March 2011) 763–767. 29. Hameed, S. A., Yuchoh, H., and Al-khateeb, W. F., A model for ensuring data confidentiality: In healthcare and medical emergency, 2011 4th Int. Conf. on Mechatronics (ICOM) (17–19 May, 2011), pp. 1–5. 1460001-15



30. Huemer, D., Tjoa, A. M., Descher, M., Feilhauer, T., and Masser, P., Towards a side access free data grid resource by means of infrastructure clouds, Int. Conf. on Parallel Processing Workshops, 2009 (ICPPW '09) (22–25 September, 2009), pp. 198–205. 31. Jun, Li, Singhal, S., Swaminathan, R., and Karp, A. H., Managing data retention policies at scale, 2011 IFIP/IEEE Int. Symp. on Integrated Network Management (IM ) (23–27 May, 2011), pp. 57–64. 32. My Chau Tu, Dongil Shin, and Dongkyoo Shin, A comparative study of medical data classification methods based on decision tree and bagging algorithms, Eighth IEEE Int. Conf. on Dependable, Autonomic and Secure Computing, 2009 (DASC '09) (12–14 December, 2009), pp. 183–187. 33. Tsirogiannis, G. L., Frossyniotis, D., Stoitsis, J., Golemati, S., Stafylopatis, A., and Nikita, K. S., Classification of medical data with a robust multi-level combination scheme, Proc. 2004 IEEE Int. Joint Conf. on Neural Networks, Vol. 3 (25–29 July, 2004), pp. 2483–2487. 34. Ericson, K., Pallickara, S., and Anderson, C. W., Analyzing electroencephalograms using cloud computing techniques, 2010 IEEE Second Int. Conf. on Cloud Computing Technology and Science (CloudCom) (30 November–3 December, 2010), pp. 185–192. 35. Mark Hall, Eibe Frank, Geoffrey Holmes, Bernhard Pfahringer, Peter Reutemann, and Ian H. Witten, The WEKA Data Mining Software: An Update, SIGKDD Explorations 11(1) (2009). 36. Jun-Hai, Zhai, Su-Fang, Zhang, and Xi-Zhao, Wang, An overview of pattern classification methodologies, in Proc. of the Fifth Int. Conf. on Machine Learning and Cybernetics (2006), pp. 3222–3227. 37. Charalampos Doukas and Ilias Maglogiannis, Advanced classification and rules-based evaluation of motion, visual and biosignal data for patient fall incident detection, Artificial Intelligence Techniques for Pervasive Computing, International Journal on AI Tools (IJAIT ) (World Scientific Press), 19(2) (2010) 175–191. 38. Charalampos Doukas and Ilias Maglogiannis, Point of care monitoring using advanced location based and context-aware services, in Hospital Information Technology Magazine (HITE) 1(2) (Summer 2008) 40–42. 39. C. Doukas, I. Maglogiannis, I. Anagnostopoulos, and K. Perakis, A context-aware telemedicine platform for monitoring patients in remote area, Journal on Information Technology in Healthcare 5(4) (2007) 255–262. 40. Charalampos Doukas and Ilias Maglogiannis, Intelligent pervasive healthcare systems, in Advanced Computational Intelligence Paradigms in Healthcare – 3, Series: Studies in Computational Intelligence (Springer, Vol. 107, 2008), pp. 95–115, Hardcover, ISBN: 978-3540-77661-1. 41. Charalampos Doukas, Ilias Maglogiannis, and Kostas Karpouzis, Context-aware medical content adaptation through semantic representation and rules evaluation, presented at 3rd Int. Workshop on Semantic Media Adaptation and Personalization (SMAP2008) (Prague, Czech Republic, 15–16 December). 42. Centers for Medicare & Medicaid Services, http://cms.hhs.gov/ 43. Majd Alwan, Siddharth Dalal, David Mack, Steven W. Kell, Beverely Turner, Jon Leachtenauer, and Robin Felder, (2000) Impact of monitoring technology in assisted living: Outcome pilot, IEEE Transactions on Information Technology in Biomedicine 10(1) (January 2000) 192–198. 44. The Jelastic Cloud provider, http://www.jelastic.com 45. Eibe Frank and Ian H. Witten, Generating accurate rule sets without global optimization, in Fifteenth Int. Conf. on Machine Learning (1998), pp. 144–151. 46. Ron Kohavi, Scaling up the accuracy of naive-Bayes classifiers: A decision-tree hybrid, in Second Int. Conf. on Knowledge Discovery and Data Mining (1996), pp. 202–207.

1460001-16



47. le Cessie, S. and van Houwelingen, J. C., Ridge estimators in logistic regression, Applied Statistics 41(1) (1992) 191–201. 48. J. Platt, Fast training of support vector machines using sequential minimal optimization, in eds. B. Schoelkopf, C. Burges, and A. Smola, Advances in Kernel Methods — Support Vector Learning (1998). 49. John G. Cleary and Leonard E., Trigg: K*: An instance-based learner using an entropic distance measure, in 12th Int. Conf. on Machine Learning (1995), pp. 108–114. 50. E. Frank, Y. Wang, S. Inglis, G. Holmes, and I. H. Witten, Using model trees for classification, Machine Learning 32(1) (1998) 6376.

1460001-17