Medical Platforms

Giovanni De Micheli, EPFL UltraSoundToGoPierre-André Farine, EPFL BodyPoweredSenSEUrsula Keller, ETHZ MIXSEL IIErnst Meyer, UniBas PATLiSci IIKlaas Prüssmann, ETHZ WearableMRIChair: Heinrich Meyr, University of

Aachen

Giovanni De Micheli, EPFL

UltraSoundToGo

High performance portable 3D ultrasound platform

Giovanni De Micheli - EPFLLuca Benini – ETHZ

Jean-Yves Meuwly - CHUVJoseph Sifakis - EPFLLothar Thiele - ETHZ

Jean-Philippe Thiran – EPFL

Motivation

Ultrasound (US) imaging is a common and non-invasive medical technique

Several US manufacturers Typically a large apparatus Some portable systems

Limitations US operation requires sonologist in place Portable systems have limited processing capabilities Commercial systems are not flexible and closed Sw

4

Objectives

Medical breakthroughs: Portable imaging device with capabilities of current large,

expensive and stationary systems Enable 3D imaging:

- Support for telemedicine: sonologist/doctor can be remotely located

System engineering goal: Parallel scalable US computing system at mobile point of care Leverage programmable hardware and parallel computation

Multi-scale system engineering Bridge the gaps among sensing, processing and visualization

through a careful choice and composition of resources5

Grand challenges for parallel scalable computing

ultra soundapplication mapping

Achieve energy-scalable processing, storage and on-chip communication For portability and form factor

Achieve provably-correct algorithmic and software implementations For safety and certification purposes

6

Tasks and challenges

Provide and efficient and safe mapping of a computational flow onto: Programmable hardware in FPGAs Manycore ASIC processor

Determine computational kernels Provide acceleration for kernels Explore power consumption/performance spectrum Investigate memory bandwidth limitations

Interface computational engine to Probes and analog circuitry Local or remote display engine Numerical analysis environments (Matlab)

7

Pierre-André Farine, EPFL

BodyPoweredSenSE

BPS – BodyPoweredSenSE:Body Powered Sensors by Scavenging Energy - Wearable I.T. for Zero Power Medical ApplicationsP.-A. Farine (ESPLAB), J. Gutknecht (NSG), T. Gross (LST),C. Hierold (MNS), D. Atienza (ESL), D. Briand (SAMLAB),

M. Knyazeva (LREN), G. Wohlrab (KISPI)

RTD 2013

BPS Project Challengeso Improve Health Care through Smart, Convenient Wearable

Sensors

o Advance Human Energy Harvesting towards a “fit and forget” and “no recharge” goal.

o Target Energy Optimisation and Usage (ETHZ 2x , EPFL 3x)

• Optimised Power Conversion from the Energy Harvesting EH sources to storage (MNS, SAM, ESP)

• ULTRA Low energy ASIC fabrication in Microcontroller design (ESL)• Self Optimising Firmware maximising the results/energy ratio (NSG, LST)

o Apply the technology in Real and Demanding Clinical Use cases (KISPI, CHUV), ensuring fitness for purpose

• Epilepsy in children (KISPI), Alzheimer’s disease in the elderly (CHUV)

o Involve Industry (ByElement, Exploris, PFIM) in multi-disciplinary IT research

LausanneMay 20th, 2014

1

BPS

BPS Wearableso EEG Baseball Cap (Brain Sensing)

• 24 Electrodes at 512 to1024 sample rates• Primarily Thermal/Solar• Role is to Intelligently Record & Compress• 1.5mW to 3mW

o ECG Chest band (Affect Sensing)• 3 Electrodes• Primarily Kinetic/Thermal• Compute HR and HRV and Record• 0.5mW to1mW

o PEM: Patient Environment Monitor (Context Sensing)• A specific clinical requirement from partners • Used to replace self reporting which is ineffective in elderly persons with

dementia and young children, provides clinicians with a hint to the context around the time of an incident recorded by the EEG and ECG sensor

• Information handled by clinicians, considered equivalent to older paper and pen diaries in terms of privacy, but more accurate and temporally superior

• Motion, Rough location, Audio (encrypted or only features stored), Video (image features or heavily pixelated images)

• Worn on the wrist as a bracelet / 5.0m to 10.0mW: Solar powered2

BPSLausanneMay 20th, 2014

Use of Energy Harvesting

(10’000µW/cm2)

(20µW/cm2)

(4µW/cm2)

Processing Systems• Sensors platform perform both physiological computations as

well as energy aware decisions to maximise utility of results (ETHZ NSG)

• Firmware executes on a specially fabricated low power ASIC that interfaces the power conditioning as well as the physiological sensor circuits (EPFL ESL)

• HomePC user friendly interface (EXPLORIS) allows gathering and sending of data to a processor farm where the clinical algorithms are managed by the clinicians (KISPI, LREN, PFIM)

• User Needs and Usability drives the overall design (BYELEMENT) UL

POWERASIC

ADAPTIVE ENERGY AWARE SOFTWARE

USER INTERFACE HOMEPC

MULTICORE DATA PROCESSING

CLINICAL OVERSIGHT 3

BPSLausanneMay 20th, 2014

Clinical Dimension• ALZHEIMER’S disease (LREN, CHUV)

o Early DIAGNOSIS using accessible EEG is critical to drug treatment efficacy

o Novel attempt to use 24 electrodes over a longer period rather than 128 Electrodes over short periods in hospital

o EEG Raw data compressed and recorded o Algorithms applied to the data:- multivariate and entropy analysis

requiring networked high end machineso Data from ECG and PEM used to improve prediction accuracy by fusing

results, combining brain, emotional and environmental factor analysis• EPILEPSY in children (KISPI)

o EEG must be made more accessible to screen for epilepsyo Todays manual methods must be enhanced with automatic brain wave

aberration detection o ECG and PEM are important to understand the context of an epileptic fit

• Evaluations o Trials will be conducted using the sensors on elderly with mild cognitive

impairment and children already diagnosed with EP

4

BPSLausanneMay 20th, 2014

Ursula Keller, ETHZ

MIXSEL II

ETH ZurichUltrafast Laser Physics

Ursula KellerUltrafast Laser Physics, Physics Department, ETH Zurich

Thomas SüdmeyerTime and Frequency Lab., Physics Department, University of Neuchatel

Gabor CsucsLight Microscopy and Screening Center (LMSC), ETH Zurich

Jacques MorelFederal Office of Metrology, METAS

Deran MaasCorporate Research, ABB Switzerland

MIXSEL IINovel semiconductor disk lasers for

biomedical and metrology applications

MIXSELModelocked Integrated eXternal-cavity Surface Emitting Laser

MIXSEL project continued with MIXSEL IIMIXSEL II project (7/13 – 10/13): What it’s about ...to develop high-power ultrafast semiconductor lasers and demonstrate their use in frequency metrology, white light generation and biomedical imaging

D. J. H. C. Maas et al., Appl. Phys. B 88, 493, 2007

gain

loss

Semiconductor Saturable Absorber Mirror

SESAM

gainloss

SESAM modelocked VECSELVertical External Cavity Surface Emitting Laser

MIXSEL II: a multidisciplinary approach

Applications semiconductor lasers

Telecom & Datacom Interconnects Optical Clocking

Frequency comb

Multi-photon imaging

First femtosecond MIXSEL pulse duration: 620 fs

output power: 101 mW

repetition rate: 4.8 GHz

center wavelength: 968 nmTBP: 0.57 sech2

MIXSELModelocked Integrated External-Cavity Surface Emitting Laser

M. Mangold, V. J. Wittwer, C. A. Zaugg, S. M. Link, M. Golling, B. W. Tilma, U. Keller, Optics Express 21, 24904, 2013

MIXSEL prototype for noise characterization

M. Mangold, S. M. Link, A. Klenner, V. J. Wittwer, C. A. Zaugg, M. Golling, B. W. Tilma, U. Keller IEEE Phot. J. 5, 1500309 (2014)

MIXSELModelocked Integrated External-Cavity Surface Emitting Laser

pulse duration: 16.7 psoutput power: 701 mWrepetition rate: 2.0 GHzcenter wavelength: 967 nm

free running: 145 fs [100 Hz - 100 MHz]stabilized: 32 fs [100 Hz - 100 MHz]

➠ Comparable to performance of DPSSL

Pulse repetition-rate tuning of MIXSEL

• repetition rate-tuning from 5 GHz to 101 GHz with single MIXSEL structure

• watt-level operation up to 15 GHz

• femtosecond operation at 60 GHz and 101 GHz

M. Mangold, C. A. Zaugg, S. M. Link, M. Golling, B. W. Tilma, U. KellerOptics Express 22, 6099, 2014

Ernst Meyer, UniBas

PATLiSci II

PATLiSci IIProbe Array Technology for Life Science

(Rapid Sensing of Cancer)

R. LimM. Loparic

Uni BaselE-Mapping

R. ZanettiUniversity

Hospital BaselOncology

E. Meyer (PI) Uni Basel

Cantilever sensors

K. GlatzUniversity

Hospital BaselPathology

H. HeinzelmannCSEM Probe array

technologiesN. de Rooij, S. GautschEPFL-IMT, MEMS design & fab

P. B. HenrichUniversity

Hospital BaselOphtalmology

U. MatterR. Sum

Nanosurf AG

End-Users:

Industrial partner:

palpation indentation

macrometer scale nanometer scale

The tip touches and feels themolecularstructures

The examiner touches and feels thepatient`s body

Force microscopy for medical applications

Sensing of Cancer

1 map = 1024 force measurements

“soft”

“stiff”

Stiffness map

Cancer "aggressiveness" peak

Inverse correlation of nanomechanical phenotypes

1. Diagnostic value

- healthy, benign and cancer

2. Prognostic value

- forecast aggressiveness of tumour

- optimize treatment e.g. chemotherapy

Plodinec et al., Nature Nanotechnology (2012)

Parallelization for Rapid Sensing of Cancer

Aim: Reduction of the acquisition time from several hoursto minutes

Proposal: Use of arrays of cantilevers, which acquire force vs. distance curves in parallel

Histograms can be acquired in a few seconds at a selectedspot. The whole acquisition time is drastically reduced

Microfabrication technology of probe arrays already validated

Probe arrays from N. de Rooij, S. Gautsch

Rapid biomarker testsLocal biomarker tests of biopsy samples will be performedwith pre-functionalized cantilever arrays to complete the tumor diagnosis

The main focus will be on the detection HER2 gene andits proteins which gives information about the characterof the tumor and about its progress (concentration)

F. Huber et al., Nature Nanotechnology 8, 126 (2013)

cells

Indentation-typeAFM

targetadsorption(sensing)

Klaas Prüssmann, ETHZ Christoph Barmet, ETHZ

WearableMRI

Wearable MRI detector and sensor arrays

K. Prüssmann, Q. Huang, G. TrösterBiomedical Engineering / Integrated Systems / ElectronicsETH Zurich, University of Zurich

Magnetic Resonance Imaging (MRI)

• Essential tool in medical diagnostics

• Widely used in basic and clinical resarch

• Global market USD 5-10 bn / a

• Key players:

Siemens, Philips, GE, new: Samsung

LimitationsLimitationsLimitations

… but are poorly implemented as yet

• fixed size and shape efficiency

• rigid setup patient comfort, application range

• complex, bulky cabling safety

• bulky electronics number of channels

Array Detectors

have revolutionized MRI in terms of

Sensitivity & Speed

Magnetic Resonance Imaging (MRI)

LimitationsLimitationsLimitationsArray Detection for MRI

Experimental head detector arraysCourtesy Prof. Larry Wald, Massachusetts Institute of

Technology

LimitationsLimitationsLimitationsWearable MRI Detectors

Proposal: Wearable MRI detectors

• flexible, elastic

• integrated on-detector receiver

• optical signal transmission

• optical power supply

scalability sensitivity freedom of motion ergonomics safety

LimitationsLimitationsLimitationsWearable MRI Detectors

Key challenges:

• detector coils: wearability & sensitivity

• integrated receiver circuit, high SNR, dynamic range

• harsh electromagnetic environment

• extreme clocking requirements

• power requirements (< 1W / channel), optical supply

• introduction to medical application

Medical Platforms

Giovanni De Micheli, EPFL UltraSoundToGoPierre-André Farine, EPFL BodyPoweredSenSEUrsula Keller, ETHZ MIXSEL IIErnst Meyer, UniBas PATLiSci IIKlaas Prüssmann, ETHZ WearableMRIChair: Heinrich Meyr, University of

Aachen