Wireless body area networksfor ambulatory health monitoring

with prototype demonstration

August 25, 2006

Chris A. Ottochrisaotto@yahoo.com

Electrical and Computer Engineering Department University of Alabama in Huntsville

Outline

� Introduction

� Proposed Solution

� WBAN Prototype

� WBAN Wireless Communications

� Embedded Software

� Personal Server Application

� Demonstration

� Conclusions

Motivation

� Healthcare is the Largest Segment of US Economy� $1.8 Trillion in 2004 (15% of GDP)

� $200 Million Informal Care Givers

� $4178 per capita (50% more than the next nearest nation)- US is less than 10th in life expectancy

- US is 26th in infant mortality rates

� Pending Crisis� Retiring Baby Boomers

� Elderly is the Largest Growing Age Group

� 45 million Uninsured

Motivation

� Current Healthcare Systems are Centralized, Focused on Reacting to Illness

� We are in need of Distributed Systems, Focused on Proactive Wellness Management

Healthcare Spending by Category

Drugs

15%

Diagnostic

4%

Treatments,

Hospital Stays,

and Rehabilitation

81%

Tobacco 18.1%

Poor diet and physical inactivity 16.0%

Alcohol consumption 3.5%

Microbial agents 3.1%

Toxic agents 2.3%

Motor vehicles 1.8%

Firearms 1.2%

Causes of Death in the US in 2000

Ambulatory Health Monitoring

� Wearable Systems For Health Monitoring

�Close monitoring of Vital Signs

�Quantitative Feedback

�Computer Assisted Rehabilitation

� Holter Monitors

�Data Recorders ( < 24 hours)

�Post-session Analysis

� Telemedicine Systems

CardioLabs

� Ambulatory Heart Monitoring

� Event-based Recorders

�Loop Recorders (32 min)

�Patient Presses “event”button during episode

� Data Extracted for Post-Analysis

CardioNet

� Mobile Cardiac Outpatient Telemetry (MCOT)

� Wireless Sensor and Wireless Monitor

� Heartbeat by Heartbeat Monitoring

� Arrythmia event detection

Heart Rate Monitors

� Polar Electro, Suunto, Timex, Reebok,…

� Integration into Fitness Equipment

� Real-time Heart rate

� Some “Journal” capabilities

ActiWatch

� Cambridge Neurotechnology

� Actigraphy

� Single Axis Accelerometer

� Clinical Research� Sleep / Wake Patterns

� Sleep Disorders

� Periodic Leg Movement (PLMS)

� Infant Monitoring

Body Media (bodyBugg)

� Multi-modal sensing� Single axis-accelerometer (motion)� (2) Temperature Sensors (heat flux)� Galvanic Skin Response (GSR)

� Upload Data using USB� Calorie Consumption Estimation

� Proprietary Algorithms � Clinically tested

CodeBlue

� Harvard University, Boston University School of Management, and Boston Medical Center, 10Blade (start-up)

� Real-time Triage, Disaster Relief� Pulse Oximeters, ECG, accelerometers

Outline

� Introduction

� Proposed Solution

� WBAN Prototype

� WBAN Wireless Communications

� Embedded Software

� Personal Server Application

� Demonstration

� Conclusions

Proposed Solution

� Wireless Body Area Network (WBAN) for Ambulatory Health Monitoring

� Mobility

� Increased Quality of Life

� Multimodal Physiological Monitoring

� Hierarchical Multi-tier Telemedicine System

System Architecture

User2

nc

InternetInternet

Tier 1:WBAN

A A

AE

Tier 2:PS

Tier 3:MS

WBAN

WWAN(GPRS)

Emergency

Informal

caregiver

Healthcare

provider

UserN

Medical

Server

WWANWLAN

WLAN(Wi-Fi)

…

User2User1 UserN…

ZigBee or

Bluetooth

User1

Outline

� Introduction

� Proposed Solution

� WBAN Prototype

� WBAN Wireless Communications

� Embedded Software

� Personal Server Application

� Demonstration

� Conclusions

WBAN Prototype - Goals

� Implement a Working WBAN� Activity Monitor / Motion Sensor

� ECG Sensor

� Network Coordinator

� Personal Server

� Explore Challenges � Sensor Fusion

� On-Sensor Processing

� Communications

� Power Efficiency

Hardware Architecture - Tmote Sky

TI MSP430 Microcontroller

10

-pin

h

eade

r6-p

inhe

ade

r

UART[0] 2

I2C[0] 2

GPIO 2

ADC[0-3] 2

ADC[6-7] 2

GPIO 2

UserINT

Reset

SPI[0]

3

ST Flash1MB

CC2420 Radio2.4 GHz

IEEE 802.15.4 Compliant

SPI SPI

UART[1]

FT232MBUSB controller

RX/TX

2

CSCS

P4.4P4.2TemperatureSensor

Humidity Sensor

LightSensor

JTAG 8-pinIDC header

7JTAG

JTAG

ADC[5]

ADC[4]

I/O

TI MSP430 Microcontroller

10

-pin

h

eade

r10

-pin

h

eade

r6-p

inhe

ade

r6-p

inhe

ade

r

UART[0] 2

I2C[0] 2

GPIO 2

ADC[0-3] 2

ADC[6-7] 2

GPIO 2

UserINT

Reset

SPI[0]

3

ST Flash1MB

CC2420 Radio2.4 GHz

IEEE 802.15.4 Compliant

SPI SPI

UART[1]

FT232MBUSB controller

RX/TX

FT232MBUSB controller

RX/TX

2

CSCS

P4.4P4.2TemperatureSensor

Humidity Sensor

LightSensor

JTAG 8-pinIDC headerJTAG 8-pinIDC header

7JTAG

JTAG

ADC[5]

ADC[4]

I/O

Hardware Architecture - ISPM

� MSP430F1232 � 32KHz (ACLK) / ~4.6MHz DCO Clock (MCLK,SMCLK)

� 8KB Flash / 256B RAM

� Two (2) ADXL202 Accelerometers

� Two Analog ECG Channels

� 10-pin Tmote Sky Header

ADXL202

Accelerometer

ADXL202

Accelerometer

ECG Signal

Conditioning

TI

MSP430F1232

CC2420

(ZigBee)

Flash

USB Interface

TI

MSP430F149

TelosISPM

ECG electrodes

ADXL202

Accelerometer

ADXL202

Accelerometer

ECG Signal

Conditioning

TI

MSP430F1232

CC2420

(ZigBee)

Flash

USB Interface

TI

MSP430F149

TelosISPM

ECG electrodes

Tmote Sky

x

y

z

ActiS Sensor Nodes

� 3 axis motion detection� Step detection and gait analysis� Activity induced

Energy Estimation (AEE)

Intelligent Signal Processing Modules (ISPM)

Polar Heart Rate Monitor

ISPM for Activity / ECG

Single chip 3-axis Accelerometer

Outline

� Introduction

� Proposed Solution

� WBAN Prototype

� WBAN Wireless Communications

� Embedded Software

� Personal Server Application

� Demonstration

� Conclusions

IEEE 802.15.4 and ZigBee

� IEEE 802.15.4� LR-WPAN Access

� 250 Kbps

� CSMA-CA

� ZigBee� Network

� Application and Application Sublayer

� Security

� Star Network Topology

Power Efficient TDMA

Super Cycle

Beacon ……Slot#3Slot#2Slot#1Beacon Beacon ……Slot#3Slot#2Slot#1Beacon

Listen Tr. Listen

Time50 ms 100 ms 150 ms 1000 ms

� >90% Sensor Power from Radio� Significant Power Savings From Disabling Radio� Timeslots for Communication

� Distributed Events � Concentrated Bursts� Allows Radio to be disabled

� Extended Battery Life / Lower Weight

TDMA Means Low Power

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50

5

10

15

20

25

Time [sec]

I [m

A]

0.65 0.7 0.75 0.8 0.85 0.90

5

10

15

20

Time [sec]

I [m

A]

Listen Transmit

Idle

superframe

� Deterministic RF timeslots

� Radio can be disabled

� Extend battery life

5 10 15 20 25 30 35 400

5

10

15

20

25

30

35

40

45

50

Channel utilization [% ]

2xAA Alkaline batteries (50gr)

Li-Ion battery (22gr) Li-Ion iPod mini battery (6gr)

Actis Processed ACC Signals

Actis Raw ACC Signals

Battery Life

Days

Typical Message Flow

NC

ACTIS_CFG_STARTMSG

ACK

ACTIS_EVENT_MASKMSGevent = HRV, R Peak

ACTIS_CALMSG

ACK

Calibration Time

ACK

Session Started

eActiS (ECG)Sensor

Personal

Server

ACTIS_EVENT_MSGevent = R Peak

Heartbeat (R peak)Detected

TIME

NC

ACTIS_CFG_STARTMSG

ACK

ACTIS_EVENT_MASKMSGevent = HRV, R Peak

ACTIS_CALMSG

ACK

Calibration Time

ACK

Session Started

eActiS (ECG)Sensor

Personal

Server

ACTIS_EVENT_MSGevent = R Peak

Heartbeat (R peak)Detected

TIME

Outline

� Introduction

� Proposed Solution

� WBAN Prototype

� WBAN Wireless Communications

� Embedded Software

� Personal Server Application

� Demonstration

� Conclusions

Embedded Software - TinyOS

� Lightweight

� Open Source

� Configuration Defined at Compile Time

� nesC (for Networked Embedded Sensors)� Component model

� Task support

� Split-phase Operation (command / event model)

Network Coordinator

� WBAN Access for Personal Server

� Network Channel Management

� Distribute Global Timing

Network Coordinator

Time Synchronization - Motivation

� TDMA

�Efficient Sharing of Communication Channels

�Timeslot Assignments

�Beacon Prediction (Maximize Radio off)

� Correlating Intra-WBAN events

�Relative timing is important

�Synchronizing start time

Time Synchronization Protocol

� Flooding Time Synchronization Protocol (FTSP)

� MAC Layer Time Stamping

� Skew Compensation with Linear Regression

Propagation

MAC Protocol DataLengthSFDPreamble

SFD � Capture Timer

Timestamp

MAC Protocol DataLengthSFDPreamble TimestampBeacon

Reception

SFD � Capture Timer

TimeSyncM

(FTSP Algorithm)

Process Send

Beacon Transmission

Propagation

MAC Protocol DataLengthSFDPreamble MAC Protocol DataLengthSFDPreamble

SFD � Capture Timer

TimestampTimestamp

MAC Protocol DataLengthSFDPreamble MAC Protocol DataLengthSFDPreamble TimestampTimestampBeacon

Reception

SFD � Capture Timer

TimeSyncM

(FTSP Algorithm)

Process Send

Beacon Transmission

Sensor Nodes

� Collect Physiological Data

� Process Data (Feature Extraction)

� Event Management

� Resource Management

� WBAN Communications

Sensor Nodes

Feature Extraction – Step Detection

46 46.5 47 47.5 48 48.5 49 49.5 50 50.5 51

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

Accelerometer based step recognition

Time[s]

Vy[m/s]Alpha[rad]accY[g]StepFoot switch

Step detected

Acc Signal HP FilterAC Signal

∑

-+ DC Signal

Orientation

Energy

Estimator

Feature Extraction – Activity Estimation

dtaACaACaACAEEt

tzyxt ⋅++= ∫

−δ

222 )()()(

10 15 20 250

2

4

6

8

10x 10

5

10 15 20 2550

100

150

Slowwalking Running

Fastwalking SittingSitting

Sta-nding

AEE

Heart rateTime [min]

Time [min]

Feature Extraction – AEE and Heart Rate

Time [min]

Out of range Burst uploads

Event Management

AEE

HR

WBANPkts

Event Management

105 105.2 105.4 105.6 105.8 106 106.2 106.4 106.6 106.8 1070

1000

2000

105 105.2 105.4 105.6 105.8 106 106.2 106.4 106.6 106.8 1070.5

1

1.5x 10

4

105 105.2 105.4 105.6 105.8 106 106.2 106.4 106.6 106.8 1071000

2000

3000

4000

accX

accY

accZ

Heart Beat

Event Message

with Timestamp

…

Beacon

Message

…

Heart Beat Step Heart Beat Step

Frame i-1

Motion

Sensor

(TS2)

ECG

Sensor

(TS1)

TS1 TS2NC TS3

Frame i

Beacon

Message

TS1 TS2NC TS3

105 105.2 105.4 105.6 105.8 106 106.2 106.4 106.6 106.8 1070

1000

2000

105 105.2 105.4 105.6 105.8 106 106.2 106.4 106.6 106.8 1070.5

1

1.5x 10

4

105 105.2 105.4 105.6 105.8 106 106.2 106.4 106.6 106.8 1071000

2000

3000

4000

accX

accY

accZ

Heart Beat

Event Message

with Timestamp

…

Beacon

Message

…

Heart Beat Step Heart Beat Step

Frame i-1

Motion

Sensor

(TS2)

ECG

Sensor

(TS1)

TS1 TS2NC TS3TS1 TS2NC TS3

Frame i

Beacon

Message

TS1 TS2NC TS3TS1 TS2NC TS3

Resource Management – USART0

USART0

RadioC

olle

ct

Sam

ple

Rx

Ra

dio

Time

50 ms

SPI SPI

100 ms 150 ms

SPI SPI

Co

llect

Sam

ple

Co

llect

Sam

ple

Co

llect

Sam

ple

Co

llect

Sam

ple

Co

llect

Sam

ple

UART

Tx

Ra

dio

Tx

Ra

dio

Tx

Ra

dio

Tx

Ra

dio

Scheduled

Beacon

25ms

ScheduledTimeslot

SPI SPIF

lash

Op

s

Reserved

For Flash

Outline

� Introduction

� Proposed Solution

� WBAN Prototype

� WBAN Wireless Communications

� Embedded Software

� Personal Server Application

� Demonstration

� Conclusions

Personal Health Server

� Sensor Node Identification and Configuration

� Sensor Fusion

� Session File Management

� Graphical User Interface (GUI)

Health Server – Configuration

Health Server – Real-time Display

Outline

� Introduction

� Proposed Solution

� WBAN Prototype

� WBAN Wireless Communications

� Embedded Software

� Personal Server Application

� Demonstration

� Conclusions

Outline

� Introduction

� Proposed Solution

� WBAN Prototype

� WBAN Wireless Communications

� Embedded Software

� Personal Server Application

� Demonstration

� Conclusions

Conclusions

� Working WBAN Prototype� Custom-designed ISPM daughter cards� Off-the-shelf wireless sensor platforms� Standard IEEE 802.15.4 communications

� Original Solutions� Health monitoring specific, Power Efficient TDMA Scheme� WBAN communication protocol� On-sensor real-time signal processing

� Promising Technology� Optimal Treatment of Disease Rehabilitation� Low Cost Early Diagnosis� Encourages Wellness Management� Improved Mobility� Low Weight (Increased Compliance)

Future Research

� Validation� Security

� Hardware encryption of wireless communications� Standard security mechanisms from the personal server to the

upper levels of hierarchy

� Signal Processing� Discern category of activity� Improved Step Detection� Improve ECG Analysis

� Upper Tier Development� Integration into Electronic Medical Records (EMR)

Acknowledgements

� Dr. Emil Jovanov, Dr. Alex Milenkovic

� Elise Haley, Corey Sanders, Reggie McMurtrey, John Gober

� University of Alabama in Huntsville