1

1

Portable Telemedical MonitoringUsing Wireless Sensors

on the Edge of the Internet

University of Alabama in HuntsvilleElectrical and Computer Engineering Dept.

Huntsville, AL 35899

Emil Jovanov

email: jovanov@ece.uah.edu

2

Intelligent monitors

• Sudden collapse victims, result ofl circulatoryl hypoxemicl traumatic arrest

• estimated mortality 350,000 lives/year• economic cost of trauma related injuries

$400 billion / year (NIH PA-01-054)

2

3

Intelligent Personal Monitors

Personal server

ECG & breathing

sensors

EEG sensors

Movementsensors

Movement

sensors

4

Can we make it without wires?

• Wireless infrastructurel 700 million subscriber units by 2002

• Mobile computingl system on chip

• Intelligent wireless sensors• Intelligent health monitors / warning devices• Personal Area Network (PAN)

l body network / intelligent clothesl wireless personal area networkl hierarchical processing

3

5

Solution?

Wireless Personal Area Network ofIntelligent Sensors

+

Hierarchical Digital Signal Processing

6

Problems of existing systems?

• Development environmentl custom VLSI, assembler?

• Resources for sophisticated real-timeprocessingl memoryl speed

• Price

4

7

Wireless Personal Area Network

• Wireless network of intelligent sensors• Wireless Intelligent Sensor (WISE)

Internet

Telemedical server

PANPersonal networkserver

WISE

• Piconetwork• Sensors

l EEGl ECGl breathingl movement

8

Wireless Intelligent Sensor WISE• Low power microcontroller TI430F149

l 16-bit RISC architecture, 60KB flash, 2KB RAMl ultra-low power consumption (400 µA in active

mode, as low as 0.8 µA in standby mode)l 8 channels (12+2 bit A/D)

• Wireless transceiverl LINX RF transceiver 916MHzl 33.6 Kbps data transfer ratel adjustable power/range

• Biomedical amplifierl Teledyne TETMD A110 (ECG/EEG amplifier)l custom amplifiers

5

9

WISE Architecture

microcontrollermicrocontroller

wireless wireless transceivertransceiver

ECG amplifierECG amplifier

ECG electrodesECG electrodes

10

WISE Ver. 1.2

microcontrollermicrocontroller

wireless wireless transceivertransceiver

ECG amplifierECG amplifier

6

Hierarchical Signal ProcessingInternet

Telemedicalserver

Personal Area Network

Personal networkserver

WISE

10 MIPS 10’s mWI

100MIPS 100’s mWII

500 MIPS 100’s WIII

IV

12

Wireless PAN configuration

Personal WISENetwork

Personal server

ECG & breathingsensors

EEG sensors

movement sensors

Movement sensors

Internet

movementsensors

7

13

Hierarchical DSP System Design

• Constraint programmingl set of constraints

• Process scheduling• Performance evaluation• Energy profiling• Iterative process

14

Process scheduling issues

• System parametersl execution timel energy per processl communication

• Constraint based scheduling

8

15

System Implementation

…

Trace

collectionMODULES

Module_n

Module_2

Module_1

Compiler_n

Compiler_2

Compiler_1

SYSTEM

Ln

L2

L1

PreprocessorConfiguration

file

Ln

L2

L1

Constraints

file

Constraint

solver

Trace

file

Application

source code

ΙΙΙ µC

I Workstation

II DSP

Hierarchical process scheduling

9

Environment for dynamic energy profiling

Jovanov, et al., “Jovanov, et al., “Environment for Energy Profiling of DSP Applications”, ICSPAT 2000, Dallas, TX.Environment for Energy Profiling of DSP Applications”, ICSPAT 2000, Dallas, TX.

DSP Application Profiling Example

CF ActiveCF ReadCPU Active

10

Time

Price

Power

Exploring the design space

DSPDSP

µµµµCC

PowerPower

ProcProc. Time. Time

HierarchyHierarchy

BatteryBatterylifelife

samplingsamplingcyclecycle

20

Examples

• ECG analysis (ischemic event monitor)• Activity monitor• Breathing monitor• Civil Disaster Data Acquisition Device

11

21

ECG Analysis

22

• Sophisticated analysis in real timel High performance/low power DSP

• Warnings (sometimes life saving)• Typical ECG processing algorithms

l Initialization (thresholds, polarity, gain control)l Filtering (band pass filters, notch filters)l QRS complex detectionl Heart rate variability processingl Baseline correction

• Robust algorithms

ECG analysis

12

23

ECG Holter Development Environment

Keyboard Input

ECGamplifier

AD

LCD Output

TMS320C5410DSP

DA

Compact FlashInterface

Programmable Logic

CompactFlash

Memory Card

AIC

SerialInterface

PC/Workstation

Speaker/Headphones

512Kx16Flash

Microphone

24

Hierarchical processing issues

• Data mining on higher levels• Adaptive thresholds• Multi-sensor synergyl running/sitting?

• Environmental effects

13

25

User activity monitoring

26

Accelerometer basedWireless Intelligent Sensor

• Analog Devices ADXL 202/210 MEMS• Digital front-end of WISE

14

27

Alternative processing methods

28

Breathing sensor

15

29

Breathing sensor• Custom analog front end• Thermistor based differential breathing sensor

30

Circadian breathing rhythm analysis• Wireless link• When_available archiving• Hierarchical processing

16

31

ECG system setup and monitoring

• System setup & debugging• Visual Basic application on portable computer• WISE Gateway with serial link interface

32

Wearable physiological monitors

• ECG (heart activity)l myocardial ischemial arrythmial circadian rhythm analysis of heart rate variability

• EEG (brain activity)l epileptic seizure detectionl drowsiness detection

• Heterogenous sensors (polygraphy)l sleep apnea monitoringl physical therapy feedback for stroke victimsl new generation human-computer interfaces

17

33

Wireless PAN - Applications

• Wearable physiological monitorsl intelligent monitoring/early warningsl decrease hospitalizations & nursing visits

• Intelligent control of medicationl sensing, dosing and compliance monitoring

• Aids for disabled• Computer assisted rehabilitation

l stroke victimsl supervised heart attack rehabilitation

• Battlefield soldier monitoring• Advanced human-computer interfaces

34

Conclusion• Enabling technology for a new generation of

telemedical systems and intelligent sensors• Sensor technology

l Implantable sensors as natural extensionl glucose blood monitors, drug pumps

• Optimum drug administration• Prolonged monitoring• Portable “guardian angel”• Research issues

l resource allocationl constraint solvingl power optimal system organization

18

35

Acknowledgments

• Dejan Raskovic• John Price• John Chapman• Anthony Moore• Abhishek

Krishnamurthy

DSP ChallengeDSP Challenge Student projectsStudent projects• Lou Woods• Daniel Pritchett• Sergey Dergunov• Dejan Milutinovic

CollaboratorsCollaborators• Dr. Krishna Kavi• Dr. Tom Martin