biotelemetry and body sensors: enabling ecg monitoringbiotelemetry and body sensors: enabling ecg...
TRANSCRIPT
BiotelemetryandBody Sensors:
Enabling ECG Monitoring
ANDREW JURIK
UNIVERSITY of VIRGINIA
18 MARCH 2009
2 of 29
This presentation will proceed as follows.
• Problem Statement and Motivation• Thesis Statement• Related Work• System Architecture• Experiments• Results• Conclusions• Future Work
3 of 29
The problem: obtaining, analyzing, and transmitting an electrocardiogram signal
• ECGs can reveal information about heart disease
• Heart disease1
• Leading cause of death in the United States (in 2005, accounted for 27.1% of all U.S. deaths)
• Projected to cost more than $304.6 billion in 2009
• Telehealth2
• For congestive heart failure patients, telehomecare resulted in 60% reductionin hospital admissions, 66% decrease inED visits, and 59% cut in pharmacy utilization
1http://www.cdc.gov/heartDisease/statistics.htm2http://www.advamed.org/NR/rdonlyres/2250724C-5005-45CD-A3C9-0EC0CD3132A1/0/TelehomecarereportFNL103107.pdf
www.drsharma.ca
4 of 29
The subproblem: how to address the technical challenges
• Wearability
• System lifetime
• Accuracy / signal fidelity
• Transmission latency
• Confidentiality
• Integrity
• Availability
5 of 29
A solution requires an interdisciplinary effort.
• Software• Andrew Jurik• Prof. Alf Weaver
• Hardware• Jonathan Bolus• Steve Jocke• Stuart Wooters• Prof. Travis Blalock• Prof. Ben Calhoun
Sponsored by the National Science Foundation through the Wireless Internet Center for Advanced Technology (WICAT)
6 of 29
Thesis statement
Our Mobile Biotelemetric System (MBS) is capable ofproviding an end-to-end “control plane” in which data isexchanged bidirectionally by means of an adaptivepolicy engine. The sensor accurately detectsheartbeats and the mobile device enforces policies,based on properties of the sensor connection and thesensor information, to promote efficient cross-systemtradeoffs.
7 of 29
The goals of our system reflect the technical challenges.
• Create a “patch”-sized sensor for the chest• Operate for 24 hours before recharging• Be at least as accurate as other heartbeat detection
algorithms• Deliver data in seconds, not minutes, to the web portal• Adopt security best practices
8 of 29
Our basic approach is to use a three-tier architecture with a policy engine.
• Make ECG sensor operationadaptive to user preferences
• Use signal processing techniques to trade off fidelityand lifetime
• Leverage “Web 2.0” for remotemedical monitoring andvisualization of data
9 of 29
Several medical (and ECG) monitoring solutions have been developed.
• Academia• CodeBlue• Human++• AMON
• Industry• CardioNet• Medtronic Carelink Network• Biotronik Home Monitoring Service• Personal Care Connect
10 of 29
The ECG sensor is where it all starts.
• Discrete components on a printed circuit board• 3 snaps• MSP430 microcontroller• RN-41 Class 1 Bluetooth radio• 430 mAh lithium-ion battery
11 of 29
Heartbeat detection is an important algorithm for ECG analysis.
• Heartbeat (QRS) detection• Pan-Tompkins algorithm• Amplitude and first derivative• First derivative only• First and second derivatives• Other digital filters• Wavelets
Time (ms)
12 of 29
Operational modes permit system-level tradeoffs.
• Heartbeat• High computation cost, low communication cost
• Waveform• Low computation cost, high communication cost• Two sub-modes:
Indefinite Specified number of seconds
13 of 29
The MBS client service is the brain of the system.
• Signal processing• Chart generation• Data logging• Operational mode changing• Exports signal via web service
14 of 29
The external network and the portal provide the remote interface.
• Web services: push to/pull from database• Send/accept alert message• Change operational mode (web portal)• Retrieve latest information on
heart rates
• MySQL database to hold user information and data
• Highslide JS and XML/SWFcharts libraries
15 of 29
Initial view of web portal after authentication
16 of 29
Data streaming view of web portal
17 of 29
The policy engine harbors the core functionality of the mobile device.
• A policy is a mapping from an event to an action.
• The policy engine implements the set of policies.
• Policies can be applied to other body sensors.
18 of 29
The policy engine can be used as a security mechanism to protect mobile device data.
19 of 29
Policy events and actions define when (events) and how (actions) MBS operates.
• Events• Disconnection• Signal timeout• Low heart rate• Sensor removed• Low battery on sensor
• Actions• Change mobile device state• Send message to portal• Send message to sensor• Ignore
20 of 29
MBS can last for more than half a day without recharging.
• Sending heart rate alone is much less power-intensive than sending the full ECG waveform
• The sensor lasts for 12 hourswhen sending heart rate data
• In waveform mode, sensor requires94 mW (19.7 dBm)• Bluetooth uses 87 mW (19.4 dBm)
on average
• The PDA can last up to 18 hourswhen receiving heart rate data(without broadcast)
21 of 29
MBS possesses an accurate heartbeat detection algorithm.
• 48 half-hour ECG signals fromMIT-BIH Arrhythmia database• Gold standard with cardiologist
annotations
• Sensitivity• Proportion of true positives over
all ground truth beats (TP+FN)• Average: 99.54%
• Positive Predictivity• Proportion of true positives over
all beats detected (TP+FP)• Average: 99.79%
22 of 29
Reducing the sampling rate does not have a large impact on the accuracy of heartbeat detection.
• Benefits of reducing sampling rate:• Processor may idle for longer• Memory (and communication) requirement is relaxed
• Drawback of reducing sampling rate:• Clinical accuracy of ECG (for heart rate information alone, lower
sampling rates are generally sufficient)
Sampling Rate (Hz)
Average Sensitivity
Average Positive
Predictivity
R-R Error (ms)
1000 99.54% 99.79% 60.55500 99.48% 99.50% 91.33250 99.40% 99.72% 72.91100 99.12% 99.09% 80.95
23 of 29
The web portal displays adequate performance.
• Less than 2 seconds to update the display with heart rates of 10 people
• Can handle multiple users (default is ten) with real-time heart rate data updates
• A subset of those canhave their ECGsdisplayed periodically
24 of 29
The policies can be made robust against transient conditions.
Window Length
Grace Period (seconds)1 5 10 15
1 beat 0.998 0.9998 0.99996 1.02 beats 0.9998 0.99996 1.0 1.04 beats 0.99993 0.99998 1.0 1.08 beats 0.99997 1.0 1.0 1.0
Sampling Rate
Grace Period (seconds)1 5 10 15
100 Hz 0.998 0.9998 0.99997 1.0250 Hz 0.998 0.9998 0.99996 1.0500 Hz 0.998 0.9997 0.99997 1.01000 Hz 0.998 0.9998 0.99996 1.0
More than 99.97% of the MIT-BIH R-R intervals are between 273 ms and 2000 ms (220 bpm and 30 bpm, respectively) for up to 5 seconds when varying either the sampling rate of the ECG signal or the number of beats over which to calculate the heart rate.
25 of 29
We present an analysis of MBS in terms of potential threats.
• Assets• ECG and/or heart rate data• Web portal database• Sensor• Mobile device
• Access points• Bluetooth radio of sensor• Bluetooth radio of mobile device• 802.11 radio of mobile device• Web services
• Techniques• Bluetooth
authentication/encryption • SSL• Web portal accounts• Limited-access database account
26 of 29
Conclusions
• MBS is an effective system• Accurate• Mobile• Near real-time• Needs a low-power radio (not Bluetooth)
• A policy engine that interfaces the signal to the rest of the system is a useful means for controlling system properties• Manage tradeoff between sensor lifetime and signal fidelity
27 of 29
Future Work and Broader Implications
• The vision is for body sensors to be small and unobtrusiveso that they are barely noticeable.
• We’ll look into how more sensors (both on one person and within a group of people) can be integrated in a framework
• Can we use authentication based on an ECG signal, and perhaps other body-derived signals?
• Can energy harvesting be used to support an ECG sensor?
28 of 29
Work related to this project has been published in several venues.• S. C. Jocke, J. F. Bolus, S. N. Wooters, A. D. Jurik, A. C. Weaver, T.N. Blalock, and B.H. Calhoun. "A
2.6-µW sub-threshold mixed-signal ECG SoC," in Symposium on VLSI Circuits, Kyoto, Japan, June 2009.
• B. H. Calhoun, J. Bolus, S. Khanna, A. D. Jurik, A. C. Weaver, and T. N. Blalock. "Sub-threshold operation and cross-hierarchy design for ultra low power wearable sensors," in IEEE Int'l Symposium on Circuits and Systems (ISCAS), Taipei, Taiwan, May 2009.
• A. D. Jurik, J. F. Bolus, A. C. Weaver, B. H. Calhoun, and T. N. Blalock. "Mobile health monitoring through biotelemetry," in Int'l Conference on Body Area Networks (BodyNets), Los Angeles, CA, USA, April 2009.
• A. D. Jurik and A. C. Weaver. "Control, analysis, and visualization of body sensor streams," in Int'l Symposium on Medical Information and Communication Technology (ISMICT), Montreal, QC, Canada, February 2009.
• A. D. Jurik and A. C. Weaver. "Body area networks: wireless access to physiological data," IEEE Software, vol. 26, no. 1, January 2009.
• A. D. Jurik and A. C. Weaver. "Remote medical monitoring," IEEE Computer, vol. 41, no. 4, pp. 96-99, April 2008.
