Wearables for novel healthcare paradigms Nick Van Helleputte

Transcription

1 Wearables for novel healthcare paradigms Nick Van Helleputte R&D manager biomedical circuits & systems - imec

2 Chronic disease management

3 Chronic disease example: United states 117 million americans suffer from one or more chronic diseases. 50% of all adults

4 CHRONIC DISEASE EXAMPLE: EUROPE 33% of all adults

5 CHRONIC OBSTRUCTIVE PULMONARY DISORDER 3 rd cause of death in Europe ; deaths related to pulmonary diseases COPD is 10 th cause of death in Japan ; pneumonia (+ influenza) is 1 st cause of death in Japan TODAY THERE ARE VERY LIMITED TOOLS TO MONITOR THESE PATIENTS AT HOME 5

6 Sleep apnea medically relevant sleep apnea meaning at least five to ten breaths are missed per hour does not only make you (permanently) tired, lacking in motivation and exhausted, but has serious consequences for your health 3-7 % of middle-aged men and 2-5 % of middle-aged women suffer from sleep apnea 9 % of Japanese men and 2.8 % of Japanese women suffer from sleep apnea 6

7 Congestive heart failure 38% of all deaths in Europe are due to cardio (or cerebro) vascular disease 24 % of all deaths in Japan are due to cardio (or cerebro) vascular disease 7

8 HYPERTENSION 9-20 % of all adults and 40-60% of all elderly in Europe have hypertension 31% of all people in Japan have hypertension ONLY 12% OF ALL PATIENTS CHANGES HIS/HER LIFESTYLE

9 Chronic Kidney Disease Relevant parameters to monitor 9

10 Role of Wearables for chronic disease management Key to improving quality of life Early, affordable and reliable diagnosis Disease progression monitoring Personalized therapy and lifestyle change Need for multi-modal, ultra-low-power signal acquisition platforms sensor fusion algorithms Personalized and contextualized data analytics 10

11 Multi-modal wearables Gyroscope

12 Multi-modal wearables Heart rhythm and HRV Respiration patterns Body water composition Physical activity Core body temperature Blood pressure Blood oxygenation (SpO2) Breath analysis (CO2, NOx,...) Body sounds... Sensor fusion algorithms Artefact reduction Feature extraction and classification Personalized & learning algorithms

13 These wearables BUILD on circuit innovations inp inn n1+ n gm1 CFBIA TI1 outp V op1 TI fb1 V on1 vp fb C cm C cm TI fb DC-servo TI fb V cp V cn n1- C dm vn fb n2+ n2- ḡm2 + CFBIA TI2 outn V on1 V op1 V op1 V on1 V op2 V on2 Level shift n1+ V x1 V cp I sig V x2 Level shift n1- OUTn V cp Vcn OUTp V cp OUTn V bn Tail current reuse gm1 TI1 TI2 13

14 Next generation ASICs are key enablers Vision & Ambition multi-sensor interface embedded signal processing power management radio (data) security 14

15 Next generation ASICs are key enablers Dedicated analog front-ends Ultra-low power, high quality, robust, small analog front-end circuits Electrocardiogram (ECG) Bio-impedance (BioZ) Electroencephalogram (EEG) Galvanic skin response (GSR) Photoplethysmograph (PPG)... multi-sensor interface embedded signal processing power management radio (data) security 15

16 Next generation ASICs are key enablers Ultra-low-power signal processing multi-sensor interface embedded signal processing power management radio (data) security On the node ULP signal processing to minimize wireless power consumption Data synchronization Feature extraction and classification Motion artifact reduction Data compaction techniques Contextual awareness Personalized machine learning algorithms 16

17 Next generation ASICs are key enablers Highly efficient power management Highly power-efficient, low quiescent current power management Linear regulators for high accuracy analog blocks Switched regulators for digital with support for DVFS SIMO & capless topologies for a highly integrated solution and less external components Optimized efficiency for low average output power levels Support for novel battery technologies multi-sensor interface embedded signal processing power management radio (data) security 17

18 Next generation ASICs are key enablers Wireless connectivity multi-sensor interface embedded signal processing power management radio (data) security Wireless connectivity to leverage the power for the cloud Optimized ULP radio links with very low standby current Compatible with existing infrastructure for rapid user adoption and wide deployability 18

19 Next generation ASICs are key enablers Security as early as possible in the signal acquisition chain Medical data is sensitive! Medical devices are potentially life critical Guarantee data security through immediate on-thenode encryption Prevent tampering with the device by secure embedded software multi-sensor interface embedded signal processing power management radio (data) security 19

20 DIVERSE applications enabled by IMEC s ulp museic CHIPS MUSEIC V2 currently available for development projects Typical power (data-collection ECG + BIOZMF + PPG) Digital (@0.6V) Analog (@ 1.2V) Total 103uW 195.4uW 298.4uW Dedicated AFEs ECG PPG BioZ GSR Embedded memory 192Kbit SRAM General purpose DSP ARM Cortex M0+ HW accelerators MAU Samplerate converter timestamp LED drivers Fully battery powered Integrated PMU Fully battery powered

21 BEYOND WEARABLES: non-contact health sensing RADAR SENSING (remote) OPTICAL SENSING (remote & wearable) Sensor fusion CAPACITIVE SENSING (through textile)

22 BEYOND WEARABLES: non-contact health sensing S1 S2 S3 S4 Multi-location capacitive ECG sensing through bed linen

23 BEYOND WEARABLES: non-contact health sensing Capacitive Bio-impedance based Respiration rate + depth sensing through shirt and sweater CAPACITIVE SENSING (through textile)

24 BEYOND WEARABLES: non-contact health sensing Respiration+HR Subject 2 HR only Subject 1 Respiration+HR HR only Heart rate and respiration rate extracted from 2 meters distance using radar

25 Example: driver monitoring ECG SIGNAL ECG SIGNAL ACCELERO METER SIGNAL ACCELEROMETER SIGNAL GYROSCOPE SIGNAL GYROSCOPE SIGNAL

26 Towards disease prevention

27 Personal behavioral technology WE PERFORM DIGITAL PHENOTYPING BASED ON CONTEXTUAL AND PHYSIOLOGICAL DATA FROM LARGE-SCALE TRIALS AND CREATE NEW TOOLS FOR PERSONALIZED BEHAVIOR CHANGE Rich physiological information (120 statistical parameters) Learn behavior & habits & cravings Find patterns & triggers Give the right recommendation at the right time Ongoing trials: stress & mental health, smoking cessation, eating disorders Diverse contextual information (smart-phone data & self reported) Unsupervised learning algorithms

28 28 DIGITAL PHENOTYPE For increased PERSONALIZATION and BEHAVIOR change

29 Digital phenotyping: stress as a use case scenario WEARABLE AND CONTEXTUAL INFORMATION FOR STRESS DETECTION WEARABLES SMARTPHONE DEMOGRAPHICS CHILLBAND CHEST PATCH QUESTIONNAIRES LOCATION VOICE PHONE DATA Skin conductance (SC) Temperature Acceleration (3 dim) ECG Acceleration (3 dim) Stress Continuous Activity During questionnaires Food/beverage intake Sleep Gastro-intestinal symptoms Proximity Acceleration Step count Screen on/off Ambient light Temperature Humidity... 29

30 Sweet study stress in the work environment LARGE SCALE DATASET WITH CONTEXT AND PHYSIOLOGICAL DATA 945 SUBJECTS 11 COMPANIES 5 TB DATA 23,000 SELF- REPORTED STRESS RESPONSES 95,000 HOURS PHYSIOLOGICAL DATA AND COUNTING...

31 Stress physiology THE FLIGHT OR FIGHT RESPONSE Jvnkfood, CC BY-SA

32 A personalized stress model a combination of physiological signals and contextual information ECG HR HRV (SDNN, RMSSD, pnn50, HF, LF, ) GSR SCR SCPH SCL OPD Physiological model Smartphone Usage times Usage patterns Skin temperature Temp slope mean Temp Social contacts Number frequency Respiration Frequency Tidal volume Location Regularity Connection to stress Speech Prosody (intonation, tone, rhythm) Sound level Ambient sound Noise level features Stres s A unique personalized algorithm correlating your physiological signals (and context) to your stress Activity Intensity Frequency Context model Sleep Time Quality

33 HR feature SC feature ST feature Age as identifier of physiological stress dynamics RIGID physiological stress response INTERMEDIATE physiological stress response DYNAMIC physiological stress response Age Age Age NEGATIVE CORRELATION between AGE and the DYNAMICS of the PHYSIOLOGICAL stress response

34 HR feature SC feature ST feature Chronic depression levels as identifiers of physiological stress dynamics RIGID physiological stress response INTERMEDIATE physiological stress response DYNAMIC physiological stress response Depression level Depression level Depression level NEGATIVE CORRELATION between DEPRESSION LEVEL and the DYNAMICS of the PHYSIOLOGICAL stress response

35 35 FUTURE DIGITAL PHENOTYPING for PERSONALIZED BEHAVIOR CHANGE PERSONALIZED MODELS to detect stress Just-in-time FEEDBACK Include CONTEXT information for modeling and feedback Develop types of users or PERSONAS based on physiology

36 Summary The role of wearables for chronic diseases may become key to improve the quality of life Not only as passive data loggers But more as an active contributor to wellbeing Multi-modal, ultra-low-power signal acquisition ASICs with powerful integrated digital signal processing support is the trend in these wearables Digital filtering Machine learning Context aware / personalization Wearables will become instrumental towards active disease prevention and behavioral change 36