1 5 wearmesoc

Page 1

WearMeSoC Multi Functional Wearable Wireless Medical Monitoring Schekeb Fateh, Integrated Systems Laboratory, ETH Z端rich Based on a Multi Channel Data Acquisition and Communication Management System on a Chip Nano-Tera Annual Plenary Meeting, May 4th, 2015


Drag picture to placeholder or click icon to add

Introduction


Introduction

Challenges

Opportunities

Escalating Medical Costs

 Wireless Revolution  Advent of M2M and IoT  Smartphones and Tablets  Nano Electronics Integration [Source: Swiss Federal Statistical Office]

Aging Population

Growing need of the elderly

Silent Suffering

 Potentially Benefiting Healthcare

WearMeSoC Objective: Develop technologies essential to miniaturize wireless monitoring, and demonstrate its effectiveness in real medical use cases


Introduction

Wearable Devices 

Clinic & point-of-care devices:  Uncompromised biosignals acquisition and processing quality required  High cost and power consumption  Not portable

Wearable devices:  Need to trade off medical-grade signal quality with:  Minimal dimensions  Ultra-low power consumption & low cost  Multi-sensor support with limited on-board signal processing resources


Introduction

Current Wearable Device with Wireless Link  Biomedical data acquisition ASIC supporting:  ECG/ EMG/ EEG/ EOG …  3.2kHz signal bandwidth  DC signal tracking

 Device based on this ASIC  Portable  Medium size et al., Modular multi-sensor platform for portable and wireless medical instrumentation, BioCAS, 2014] [P. Schonle Wireless connection to monitoring display


Introduction

Modular Biomedical Data Acquisition Platform

 Development platform:  Modular  Flexible  Programmable


Drag picture to placeholder or click icon to add

VivoSoc: Intelligent Sensor Front-end with Wireless Link


VivoSoc

VivoSoc: System-On-Chip Architecture

 Dual core PULP (Parallel processing Ultra-Low Power platform)  Provides computing resources to also run communication protocol stack [F. Conti et al., Energy-efficient vision on the PULP platform for ultra-low power parallel computing, SiPS 2014]


VivoSoc

VivoSoC: Biomedical Data Acquisition System-on-Chip Area:

3.8 mm x 2.8 mm

Technology:

130 nm CMOS (tape-out: April 2015)

Architecture:

Dual core RISC with DMA, 2 kByte Cache, 32 kByte L2, 16 kByte TCDM

Front-end:

8 analog channels with 108 dB CMRR, 3.2 kHz bandwidth

Peripherals:

• • • •

2 quad SPI masters 1 SPI slave 8 GPIO, 1 JTAG 1 UART


VivoSoc

Vision of VivoSoC  Potential Application: Build intelligent biomedical platform with only 3 ASICs including RF capability


Drag picture to placeholder or click icon to add

Long-Distance Wireless Link


Wireless Link

Ultra Low Power Cellular Data Modem (GSM/EDGE/E-EDGE)

[Krรถll2014] H. Krรถll et al. An Evolved EDGE PHY ASIC Supporting Soft-Output Equalization and Rx Diversity, ESSCIRC 2014 [Krรถll2015] H. Krรถll et al. An Evolved GSM/EDGE baseband ASIC Supporting Rx Diversity, JSSC 2015


Wireless Link

Dual Antenna Cellular Data Modem Testbed with ML605 and Zedboard

-113.4 dBm sensitivity of Rx-diversity



Processing of a single 32-QAM EGPRS2-A time slot consumes 5.5 mW


Field of Applications


Drag picture to placeholder or click icon to add

Electrocardiography (ECG) and Pulse Oximetry USZ Pulmonary Department


Pulse Oximetry

Pulse Oximetry Hardware

First experiments based on discrete hardware Test of algorithms for motion artifact and ambient light suppression Crucial intermediate step towards an integrated solution

Oximetry module for the WearMeSoC development platform based on discrete components

walking and arm swinging

arm swinging

 Specification

First Oximetry ASIC:

still vertical movement

horizontal movement

Oscillating finger

Raw data of a test recording showing heavy motion artifacts for different movements


Pulse Oximetry

Ambulatory Respiratory Monitoring with Lung Disease 

40 patients with chronic obstructive pulmonary disease performed several 6MWT in Zurich (490m), Davos Clavadel (1650m) and Davos Jakobshorn (2590m)

Hypothesis: SpotSpO2 differs from values derived at the end of 6MWT by analysis of SpO2 trends (TrendSpO2) downloaded from pulse oximetry memory after 6MWT

Altitude had no influence on agreement

Visual spot readings and graphical analysis of SpO2 recordings at end of 6MWT agree well on average


Electrooculography (EOG) Fitness-to-Drive Test for the Elderly USZ Neurology Department


EOG

Mobile Eye Tracking Scene camera with synchronization LED

Vertical EOG channel

Horizontal EOG channel Live view of channels


EOG

Achievements  Sophisticated synchronization method  Optimal-polynomial 9point calibration procedure  System linearity over +/15deg

Target Eye Mean

 Robust event detection (blinks, fast eye movements, fixations) 2°


Drag picture to placeholder or click icon to add

Electroencephalography (EEG) USZ Sleep Analysis Lab


EEG

Automatic Artifact Detection in Long-Term EEG Recordings [A. Wierzbicka et al.: Sleep Disorders Center, Institute of Psychiatry and Neurology Warsaw, Poland]

 Implemented algorithms for automatic artifact detection  Identify artifact free segments for further analysis  Best to combine several algorithms


EEG

Identification of Microsleep Episodes

Microsleep episodes are characterized by a change in oscillatory activity

Microsleep episodes mostly associated with a lack of eye movements, and often occurred after brief episodes of alpha activity (~10 Hz)

Future work: Develop a method for (semi-)automatic detection of microsleep episodes


Drag picture to placeholder or click icon to add

Electromyography (EMG) Universita di Bologna Micrel Lab


EMG

Platform for Evaluation of EMG Signals and Hand Gesture Recognition using Cerebro JTAG/ SWD DIGITAL ANALOG SD CARD BLUETOOTH MODULE BLUETOOTH REGULATOR

USART PRESSURE SENSOR IMU

USB BATTERY

SENSORS REGULATOR

I2C

JTAG

SPI

ARM Cortex M4 MCU

MCU REGULATOR

SPI

CEREBRO

ELECTRODES CONNECTOR

CEREBRO DIGITAL CEREBRO ANALOG REGULATOR REGULATOR

POWER MANAGEMENT

• Cerebro interfaced with the ARM Cortex M4 to process EMG signals • Data is transferred using Bluetooth to a mobile device

http://www.touchbionics.com/


EMG

Hand Gesture Recognition [S. Benatti et al. EMG-Based Hand Gesture Recognition with Flexible Analog Front End, BioCAS 2014]

EMG Controlled Hand Prosthesis

92% accuracy in estimation of 7 gestures Digitally-controlled AFE


Drag picture to placeholder or click icon to add

[Center for Neuroprosthetics, EPFL | STI | IMT/IBI | LSBI]

Ultra Low Power and Implantable Biomedical Device GlaxoSmithKline (GSK)


Implantable Devices

Ultra Miniaturized and Implantable Device

Size of miniaturized and folded device around 0.8 cm3

Future Goal: use VivoSoC and target a size below 0.5 cm3


Modular Device versus Implantable Device Size: 312 cm3

Size: 0.8 cm3


Implantable Devices

Implantable Device Interfacing Nerves

Implantable Device

 First recording of nerve activity expected in June

Interfacing nerves

 Feedback including stimulation  Miniaturized device in fabrication  Electrodes already in test phase [Center for Neuroprosthetics, EPFL | STI | IMT/IBI | LSBI]

Electrodes


Drag picture to placeholder or click icon to add

Conclusions


Conclusions

Conclusions  Successful tests of portable device through medical research partners

 System-on-Chip for biomedical platform called VivoSoC realized

 ASIC implementation of Pulse Oximetry Hardware

 Ultra low power cellular modem implemented by industry partner

 Additional promising applications gained:  EMG based gesture recognition  Ultra-low power implantable devices

 First implantable system with size of 0.8 cm3


Drag picture to placeholder or click icon to add

Questions


Drag picture to placeholder or click icon to add

Supplementary Slides


Introduction

Wireless Portable Medical Devices Market

WEARMESOC Objective: Develop technologies essential to miniaturize wireless monitoring, and demonstrate its effectiveness in real medical use cases


VivoSoc

Healthcare Platform  User friendly  Minimum size, wearable platform  Low power  Multi purpose  Programmable  Storage capability


VivoSoc

VivoSoc: Realization of the Biomedical Platform 

Multichannel low power analog frontend

Data conversion

Signal processing

Various interfaces

Cascade-able

Software defined

Short time-tomarket


Applications

Clinical Use Cases  Post Operative Monitoring  A Research Focus at USZ  IT in Medicine Department  Both Size and Functionality Matters!

 Eye Position Monitoring  USZ Neurology Department  Important topic for safe driving by the elderly  Challenges for miniaturization

 Study of Altitude Disorder  USZ Pulmonary Department  Swiss spend much time in mountains  Important for safety in some workplaces


Applications

Clinical Use Cases  Sleep Disorder Research  Sleep Analysis Lab (USZ)  Both Size and Functionality Matters!  What would we give to get a good night’s sleep?

 EMG Controlled Hand Prosthesis  Micrel Lab (Universita’ di Bologna)  Classifier based gesture recognition  Performing of daily living movements

 Ultra Low Power Implantable Devices  Centre Hospitalier Universitaire Vaudois  Centre Neuroprosthetics EPFL  GlaxoSmithKline (GSK)


EEG

EEG in Sleep Research

 Today, EEG electrodes are connected to an amplifier beside the bed

 The wiring leads to an unnatural sleep situation

 A small wireless device would improve the experimental setup


EEG

Test of Prototype of Wearable Bio-Monitoring Device (Sleep Recording)

   

Four-channel recording of EEG for 9 hours SWA: EEG power in 0.75-4.5 Hz range EMG power in 10-40 Hz range Subject awake for 1.5 h engaged in activity and then went to bed


EOG

Next Steps & Challenges  Implementation of MIMO-mirror projector for head-centered target display

 Removal of physiological baseline drift  Implementation of photodiode for luminance measurements

 Driving Test for Elderly Drivers


Application

EOG: Comparison to other systems

Bluebox EOG

Video-based

1000Hz

30 - 250Hz

Influence of eye physiology

None

High

Operator expertise

Low

Medium - High

Pricing

Low

High

Modularity

High

None

~2deg

<1deg

Sampling rate

Spatial resolution


EOG: Next steps & challenges  Head-centered calibration Miniature laser projector for target display www.hamamatsu.com/sp/hc/osh/osh_05a_image01.jpg

 Removal of physiological drift Photodiode for luminance measurements

 Fitness-to-drive test Develop and test the elderly


Eye movements & Fitness-to-Drive test modified from: Yishu Liu – Electrooculography Based Fitness to Drive Test

Elderly license holder’s fitness to drive need to be tested regularly Current : Purely medical exam

Proposal: Computer-assisted evaluated on-road driving test

Does not reflect exact demand of driving

Reflects the exact demand of driving

Time consuming

One single test

No medical parameter can unambiguously determine the fitness to drive

Saccade eye movement

Easy to measure Indicates diseases like dementia Not typically examined in common neuropsychological tests

Yishu Liu

45


EOG

Comparison to Head-Mounted Video-Based Eye Tracking METHODS

Cereblue EOG

Video-based

Sampling rate

1000Hz

30-250Hz: dynamic eye movement parameters may be degraded

Influence of eye physiology

None

High: droopy eye lids cause data loss  Problematic in the elderly

Operator expertise

Low: Electrode placement

High: Eye detection parameters

Pricing

Low

High: up to 10.000$

Modularity

High: only 3/8 channels used  Additional components can easily be added (gyroscope, ECG,…)

None

Spatial resolution

Low: ~2deg

High: <1deg


Eye movements

Compensatory (reflexive)

Saccade network

Goal-directed (voluntary)


Electrooculograpy Corneo-retinal potential

Vertical electrodes

Horizontal electrodes


EOG-Video Synchronization modified from: Yishu Liu – Electrooculography Based Fitness to Drive Test

• Generate random sync. signal with the Bluebox ; stored as pulse sequence • Control LED in video with this sequence • 8-bit sync. sequence every 2 seconds • Synchronize by finding same sync. sequence

ch1 ch2 sync.

8 4 8 4 8 5 6 9 7 9 ......

0 0 1 1 1

Yishu Liu

49


Pulse Oximetry

Test ASIC: Transimpedance Medical Amplifier for Oximetry (TMA-O) 

3 channel TIA  Full swing:  Max. ENOB:

6 channel LED driver

Highly Partially parallel configurable

TIA + ADC

digital circ.

LED driver

 Arbitrary selection of input and output channels for up to 5 program sequences. On-chip calibration  Synchronization for parallel multi-chip operation

Future: Integration of the TMA-O front-end in VivoSoC


Pulse Oximetry

TMA-O: TIA front-end 

3 differential inputs  Full swing:  Max. ENOB: Schematic of the TIA

Correlated double sampling (CDS) for ambient light suppression.

Front-end can be used in resistive and capacitive TIA modes.  Enhanced input range

Frequency (Hz)

Ambient light filtering performance dependent on the pulse width.


Pulse Oximetry

TMA-O: LED driver  6 channel LED driver  resolution

 On-chip current calibration Schematic of the LED driver

 LED duty-cycle can be adapted at the signal condition for optimum power saving


EMG

EMG Controlled Hand Prosthesis Passive Prosthesis Active Prosthesis

Multifinger Prosthesis

- Cosmetic - Low Functionality - 1 or 2 DOF - EMG controlled

- High number of DOF - EMG controlled - Daily living movements


EMG

Comparison to MYO Armband

• Cerebro has similar performance as the MYO system


Implantable Devices

Top Layer of MicroDevice Rigid 11mm

11mm

Rigid 11mm

4-Layer Flex 3mm

PCB: 6 Layer Rigid-Flex, thickness 0.8mm


Implantable Devices

≈ 5.5 mm

Side View (Folded) of MicroDevice

≈ 12.5 mm


Supplementary Slides

Comparison to Existing Holter ECG System Mortara H3+  3 channel ECG  48h recording IMEC Holst Center  single channel ECG  7d recording WearMeSoC Platform  8 channel ECG/EEG/…  Real time monitoring  Size: about 1 cm3


VivoSoc in Operation


Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.