Upload
andreas-triantafyllidis
View
3.735
Download
1
Tags:
Embed Size (px)
DESCRIPTION
Presentation from 2nd International Conference on Pervasive Computing Technologies for Healthcare 2008 Tampere, Finland. Parts of this work have been published in Methods of Information in Medicine Journal 2008;47(3):229-34 available at: http://www.ncbi.nlm.nih.gov/pubmed/18473089 Objectives: Sensor networks constitute the backbone for the construction of personalized monitoring systems. Up to now, several sensor networks have been proposed for diverse pervasive healthcare applications, which are however characterized by a significant lack of open architectures, resulting in closed, non-interoperable, and difficult to extend solutions. In this context, we propose an open and reconfigurable Wireless Sensor Network (WSN) for pervasive health monitoring, with particular emphasis in its easy extension with additional sensors and functionality by incorporating embedded intelligence mechanisms. Methods: We consider a generic WSN architecture comprised of diverse sensor nodes (with communication and processing capabilities) and a Mobile Base Unit (MBU) operating as the gateway between the sensors and the medical personnel, formulating this way a Body Area Network (BAN). The primary focus of this work is on the intra-BAN data communication issues, adopting SensorML as the data representation mean, including the encoding of the monitoring patterns and the functionality of the sensor network. Results: In our prototype implementation two sensor nodes are emulated; one for heart rate monitoring and the other for blood glucose observations, while the MBU corresponds to a Personal Digital Assistant (PDA) device. Java 2 Micro Edition (J2ME) is used to implement both the sensor nodes and the MBU components. Intra-BAN wireless communication relies on the Bluetooth protocol. Via an adaptive user interface in the MBU, Health Professionals may specify the monitoring parameters of the WSN and define the monitoring patterns of interest in terms of rules. Conclusions: This work constitutes an essential step towards the construction of open, extensible, interoperable and intelligent WSNs for pervasive health monitoring.
Citation preview
An Open and Reconfigurable Wireless Sensor Network for Pervasive Health Monitoring
A. Triantafyllidis, V. Koutkias, I. Chouvarda andN. Maglaveras
Lab of Medical InformaticsFaculty of Medicine, Aristotle University of Thessaloniki
Thessaloniki, [email protected]
04/08/23 2nd International Conference on Pervasive Computing Technologies for Healthcare 2008
Tampere, Finland
2
Presentation Outline
• Sensor Networks today: Pros and Cons
• Our Proposal: A Framework for Interoperable and Open Biosensor Networks
• System Architecture
• Functional Attributes
• Data and System Representation
• Prototype Implementation Description
• Discussion and Future Work
04/08/23 2nd International Conference on Pervasive Computing Technologies for Healthcare 2008
Tampere, Finland
3
Sensor Networks: Trends and Recent Advances
• Miniaturization
• CPU Speed and Memory increase according to Moore’s Law
• OS Advances:– Multi-threading OS, Java Platform
• New Protocols:– Wireless Communication: ZigBee, WiBree– Data Representation: IEEE 1451,SensorML
04/08/23 2nd International Conference on Pervasive Computing Technologies for Healthcare 2008
Tampere, Finland
4
Sensor Networks Today: Challenges
• Interoperability:– applications “tight” to hardware– difficulties in achieving interoperability for different
sensor nodes and different services– custom solutions with maintenance problems
• Intelligence:– nodes usually not smart enough to do processing
tasks– accumulation of processing burden in a single central
node
• Energy:– high energy costs through continuous data transfer in
wireless links
04/08/23 2nd International Conference on Pervasive Computing Technologies for Healthcare 2008
Tampere, Finland
5
System Architecture Objectives
• Formal Sensor Descriptions: Sufficient description of the entire sensor network, including nodes and processes/services
• Openness: Plug ‘n play support in adding/removing nodes
• Reconfigurability: Sensors adaptation to application/user needs in run-time
• Requirements:- A data representation standard for communication
requirements: OGC® SensorML (http://www.opengeospatial.org/standards/sensorml/)
- Intelligence: Medical Rules
04/08/23 2nd International Conference on Pervasive Computing Technologies for Healthcare 2008
Tampere, Finland
6
SensorML: A New Standard for Sensor Description
• Sensor Node Identification (meta-data info)
• Sensor Capabilities (e.g. resolution, range, accuracy)
• Sensor Inputs (e.g. sensor data detected from environment, sensor data from another node)
• Sensor Outputs (e.g. sensor data, alerts according to specific criteria)
• Sensor Processes (e.g. calculation of a MIN derivative in a sensor data array)
• Sensor Connections (Connections of nodes inputs to outputs and vise versa)
04/08/23 2nd International Conference on Pervasive Computing Technologies for Healthcare 2008
Tampere, Finland
7
SensorML Example Document and Schema
Physical and Functional Entities are expressed as processes<sml: ProcessModel id=”PROCESS_ID”>
<sml:metadata … />
<sml:inputs … />
<sml:outputs … />
<sml:parameters … />
<sml:method xlink:href=”urn:{authority}:def:process:{processName}:{version}”/>
</sml:ProcessModel>
Schematron language can be used for strict definition of XML schema
<sch:rule context=”xpath expression”><sch:assert test=”xpath expression”> Message if incorrect </sch:assert><sch:assert … /><sch:assert … /></sch:rule><sch:rule … /><sch:rule … /></sch:pattern><sch:pattern … /><sch:pattern … /></sch:schema>
04/08/23 2nd International Conference on Pervasive Computing Technologies for Healthcare 2008
Tampere, Finland
8
Proposed System Architecture
04/08/23 2nd International Conference on Pervasive Computing Technologies for Healthcare 2008
Tampere, Finland
9
Sensor Capabilities/Modules
• Sensing: Acquirement of sensing phenomenon data
• Monitoring: Data processing according to pre-defined Medical Rules. Corresponding algorithms take place
• Communication: Communication with MBU via SensorML messages when an event is recognized
• Data Handling: Serialization and de-serialization of data
Sensor Modules
Sensing
Monitoring
Communication
Data Handling
04/08/23 2nd International Conference on Pervasive Computing Technologies for Healthcare 2008
Tampere, Finland
10
MBU (Mobile Base Unit) Capabilities
Additional capabilities with respect to Sensors:
– Data Aggregation• Compound rules containing
information about multiple biosignals (currently, support of AND/OR logical operators)
– Actions• Respond to sensor
• Generate alert for user
• Forward information to Medical Center
MBU Modules
Data Aggregation
Communication
Data Handling
04/08/23 2nd International Conference on Pervasive Computing Technologies for Healthcare 2008
Tampere, Finland
11
Medical Rules: Event and Action Oriented
Events:• Spontaneous Events
– Monitor whether a raw signal value or signal derivative, average value, maximum or minimum value etc, is above or below predefined thresholds within a time window
• Persistent Events– Event onset, Event end (currently, under development)
Actions:• Triggered in the form of messages to the MBU,
describing the event and optionally carrying data that triggered the event
04/08/23 2nd International Conference on Pervasive Computing Technologies for Healthcare 2008
Tampere, Finland
12
Communication Flow (1/4)
1. MBU receives Sensor Node’s self-description encapsulated in a SensorML message
Dynamic GUI construction in the MBU device Interoperability between MBU and Sensor Node is
achieved Ability of creating personalized Medical Rules
Send Sensor Description
Sensor MBU
04/08/23 2nd International Conference on Pervasive Computing Technologies for Healthcare 2008
Tampere, Finland
13
Communication Flow (2/4)
2. MBU sends Medical Rules info to the Sensors involved
Sensor Logic adaptation takes place Sensor Data processing according to Medical Rule Medical Rule example: Get the average value of heart
rate in a time window of X sec when average value above Y pulses/min
Send Medical Rules
Sensor MBU
04/08/23 2nd International Conference on Pervasive Computing Technologies for Healthcare 2008
Tampere, Finland
14
Communication Flow (3/4)
3. Sensor sends alert to the MBU when the rule criteria are met and the corresponding event is triggered, containing:
Sensor Data Event Start time Event Description
Send Alert
Sensor MBU
04/08/23 2nd International Conference on Pervasive Computing Technologies for Healthcare 2008
Tampere, Finland
15
Communication Flow (4/4)
4. MBU forwards information related to the triggered event to the Medical Center
Currently, asynchronous ways of communication are studied: SMS, MMS
Forward Information (SMS/MMS)
MBU Medical Center
04/08/23 2nd International Conference on Pervasive Computing Technologies for Healthcare 2008
Tampere, Finland
16
Prototype Features
• Sensors emulation as mobile terminals, e.g.:– Heart Rate Sensor– Blood Glucose Sensor
• MBU is a PocketPC - SmartPhone device (Currently, a Qtek MDA II device)
• Bluetooth is chosen as the wireless communication link:– Low-cost– Low-power– Used widely
• SensorML adoption for data exchange
04/08/23 2nd International Conference on Pervasive Computing Technologies for Healthcare 2008
Tampere, Finland
17
Prototype Implementation
• J2ME-CLDC (Java 2 Micro Edition- Connected Limited Device Configuration)– Open platform for mobile applications– Currently supported by Sun SPOT sensor platform– Supported by the majority of mobile devices
• JSR-82 Libraries (Java Specifications Request)– Interface that links Bluetooth hardware to Java– Ad-hoc device connection– Automatic service discovery
• KXML– Lightweight, non-validating pull-parser for serialization and de-
serialization of SensorML data (http://kobjects.org/kxml/)
04/08/23 2nd International Conference on Pervasive Computing Technologies for Healthcare 2008
Tampere, Finland
18
SensorML-based Sensor Self-description Example (1/2)
<!– Sensor Identification --><Sensor id="HR_id"><identification><identifierList><identifier name="shortName"/><identifier
name="longName"/><identifier name="modelNumber"/><identifier name="manufacturer"/></identifierList></identification>
<!-- Sensor general and functional capabilities --><capabilities><PropertyList><property
name="generalCapabilities"><DataGroup><component name="resolution">1.0</component>
<component name="range">0 200</component><component name="accuracy">-1 1</component></DataGroup></property>
<!-- Sensor functional capabilities for samplingRate and sizeOfWindow configuration -->
<property name="functionalCapabilities"><DataGroup><component name="samplingRateRangeConfigurationCapability">
0.001 1</component><component name="sizeOfWindowRangeConfigurationCapability">
1 100</component></DataGroup></property></PropertyList></capabilities>
04/08/23 2nd International Conference on Pervasive Computing Technologies for Healthcare 2008
Tampere, Finland
19
<!--Sensor inputs--><inputs><inputList><input name="heartRate"></input></inputList></inputs>
<!--Sensor outputs--><outputs><outputList><output name="minAlert"><DataGroup><component
name="eventTime"></component><component name ="measuredValue"></component></DataGroup></output><output name="maxAlert" .../><output name="averageAlert"
.../></outputList></outputs>
<!--Definition of the actuating procedure of an alert with MIN criteria, as a process model--><processes><processList><process name ="minAlertRuleTrigger"><inputs><inputList><input name="heartRate"></input></inputList></inputs><outputs><outputList><output name="eventTime"></output><output name="measuredValue"></output></outputList></outputs><parameters><ParameterList><parameter
name="heartRateMin"><Datagroup><component name="comparisonCriteria"/><component name="maxThreshold"/><component name="timeFrame"/></Datagroup></parameter></ParameterList></parameters></process>...</processList></processes>
<!-- System inputs to process inputs, process outputs to system outputs --><connections><ConnectionList><connection name="outputOfMinAlertRuleTriggerToSensor"><Link><source ref="minAlertRuleTrigger/outputs/measuredValue"/><destination
ref="this/outputs/minAlert/measuredValue"/></Link></connection>...</ConnectionList></connections></Sensor>
SensorML-based Sensor Self-description Example (2/2)
04/08/23 2nd International Conference on Pervasive Computing Technologies for Healthcare 2008
Tampere, Finland
20
SensorML-based Medical Rule Example Definition (1/2)
<process name =" AVG_HRGLU">
<!-- Rule process requests of sampling rate -->
<capabilities><PropertyList><property name="samplingRateHeart">1</property><property name="samplingRateGlucose">0.1</property>
</PropertyList></capabilities>
<!-- Rule process inputs -->
<inputs><InputList><input name="physiologicalPhenomena"><DataGroup><component name="heartRate"></component><component
name="bloodGlucose"></component></DataGroup></input></InputList></inputs>
<!-- Rule process outputs -->
<outputs><OutputList><output name="AVG_HRGLU" description="AVG_HRGLU DESCRIPTION"><DataGroup>
<component name="outputGivenAs">SMS</component><component name="smsText">PATIENT IN ALERT CONDITION</component>
<component name="eventTime" /><component name="measuredValue" /><component name="outputTimeFrame"/>
<component name="dt"/></DataGroup></output></OutputList></outputs>
04/08/23 2nd International Conference on Pervasive Computing Technologies for Healthcare 2008
Tampere, Finland
21
<!-- Rule process parameters of Average function -->
<parameters><ParameterList><parameter name="heartRateAverage"><Datagroup>
<component name="comparisonCriteria">greaterThan</component><component name="maxThreshold">100</component>
<component name="timeFrame">300</component></Datagroup></parameter><parameter name="relationship"><Datagroup>
<component name="logicCriteria">[OR]</component></Datagroup></parameter><parameter name="glucoseAverage">
<Datagroup><component name="comparisonCriteria">greaterThan</component>
<component name="maxThreshold">200</component><component name="timeFrameRule">1800</component>
</Datagroup></parameter></ParameterList></parameters></process>
SensorML-based Medical Rule Example Definition (2/2)
04/08/23 2nd International Conference on Pervasive Computing Technologies for Healthcare 2008
Tampere, Finland
22
MBU Demo Snapshots
Adaptive User Interface for medical rules definition through a step-by-step procedure: Parameter selection and parameter definition
04/08/23 2nd International Conference on Pervasive Computing Technologies for Healthcare 2008
Tampere, Finland
23
MBU Snapshots in PDA
Adaptive UI as seen in a Qtek MDA II device
04/08/23 2nd International Conference on Pervasive Computing Technologies for Healthcare 2008
Tampere, Finland
24
Discussion: Future Work
• Support for management of both spontaneous and persistent events:
– Recognition of repetitive events– Alert severity and prioritization
• Interconnection with Medical Center– SOA Architecture: Web Services
• Security issues:– Currently employed the built-in authentication mechanism of
Bluetooth– Elaborating on data encryption techniques on top of the
Bluetooth protocol stack as a more advanced security approach
An Open and Reconfigurable Wireless Sensor Network for Pervasive Health Monitoring
A. Triantafyllidis, V. Koutkias, I. Chouvarda and N. Maglaveras
Lab of Medical InformaticsFaculty of Medicine, Aristotle University of Thessaloniki
Thessaloniki, [email protected]
Thank you!!!