transforming epilepsy from a chronic condition towards an
play

Transforming Epilepsy from a chronic condition towards an acute one. - PowerPoint PPT Presentation

TRANSFORMING EPILEPSY FROM A CHRONIC CONDITION Sunnyvale, California - 20 Mar 2019 TOWARDS AN ACUTE ONE tinyML Summit Transforming Epilepsy from a chronic condition towards an acute one. Hans De Clercq, Byteflies Ben Vandendriessche Jonathan


  1. TRANSFORMING EPILEPSY FROM A CHRONIC CONDITION Sunnyvale, California - 20 Mar 2019 TOWARDS AN ACUTE ONE tinyML Summit Transforming Epilepsy from a chronic condition towards an acute one. Hans De Clercq, Byteflies Ben Vandendriessche Jonathan Dan Tomas Fiers Lieven Billiet March 20, 2019, Sunnyvale, California /18

  2. BYTEFLIES We enable lean development of wearable health TRANSFORMING EPILEPSY FROM A CHRONIC CONDITION We are building the most comprehensive data set of vital signs Sunnyvale, California - 20 Mar 2019 TOWARDS AN ACUTE ONE outside the hospital. We allow clinical studies of more than 10000 patients, no matter where they are. We translate data into objective, 24-7 clinical insights /18 2

  3. SENSOR DOT Any signal, any time TRANSFORMING EPILEPSY FROM A CHRONIC CONDITION Heart activity Muscle activity Sunnyvale, California - 20 Mar 2019 TOWARDS AN ACUTE ONE Respiration Skin conductance Motion Blood pulse Brain activity Eye motion /18 3

  4. EPILEPSY TRANSFORMING EPILEPSY FROM A CHRONIC CONDITION $15B 1/26 100k Yearly cost in US People affected Yearly deaths Sunnyvale, California - 20 Mar 2019 TOWARDS AN ACUTE ONE 20% 33% $1B/Yr Wrongly No treatment No improvement! diagnosed available /18 4

  5. DIFFICULT TO DIAGNOSE IN THE HOSPITAL TRANSFORMING EPILEPSY FROM A CHRONIC CONDITION Video EEG monitoring 24h to 5 days hospitalization $$$ Sunnyvale, California - 20 Mar 2019 No seizure = no data TOWARDS AN ACUTE ONE Self-reported outcome Less than 50% accuracy Time consuming Low patient adherence /18 5

  6. Your wearable epilepsy solution TRANSFORMING EPILEPSY FROM A CHRONIC CONDITION 82 9800 390 patients hours data Sunnyvale, California - 20 Mar 2019 seizures TOWARDS AN ACUTE ONE Optimized Sensor selection Seizure detection with 90% sensitivity Scalable demonstrator /18 6

  7. OLD VERSUS NEW TRANSFORMING EPILEPSY FROM A CHRONIC CONDITION 7 /18 TOWARDS AN ACUTE ONE Sunnyvale, California - 20 Mar 2019

  8. NEXT STEPS TRANSFORMING EPILEPSY FROM A CHRONIC CONDITION Multi-centered validation Largest Epilepsy dataset outside Sunnyvale, California - 20 Mar 2019 TOWARDS AN ACUTE ONE the hospital Online seizure detection Clinical application /18 8

  9. MACHINE LEARNING FOR HEALTHCARE: A MODULAR APPROACH TRANSFORMING EPILEPSY FROM A CHRONIC CONDITION More training data Sunnyvale, California - 20 Mar 2019 TOWARDS AN ACUTE ONE More interpretation Lower complexity Easier to validate /18 9

  10. FEATURE EXTRACTION EXAMPLE: GENERAL PURPOSE PEAK DETECTOR 10 TRANSFORMING EPILEPSY FROM A CHRONIC CONDITION /18 TOWARDS AN ACUTE ONE Sunnyvale, California - 20 Mar 2019

  11. GENERAL PURPOSE PEAK DETECTOR Approach 1: 1D convolutional NN TRANSFORMING EPILEPSY FROM A CHRONIC CONDITION simple CNN (2 conv layers), 1 pooling layer, and 1 fully connected output layer with per sample prediction. Sunnyvale, California - 20 Mar 2019 TOWARDS AN ACUTE ONE Approach 2: 2D convolutional NN simple CNN (4 conv layers), 1 pooling layer, and 2 fully TESTED ON 50.000 HR OF connected output layers with predictions per sub epoch (window of size QRS). SENSOR DOT DATA!! /18 11

  12. MODULAR LAYER EXAMPLE: MOTION TRACKING Approach: Principal TRANSFORMING EPILEPSY FROM A CHRONIC CONDITION component analysis Sunnyvale, California - 20 Mar 2019 TOWARDS AN ACUTE ONE No-activity Activity Tonic Tonic-Clonic /18 12

  13. HARDWARE SPECIFICATIONS TRANSFORMING EPILEPSY FROM A CHRONIC CONDITION ECG Dot Motion Dot Epilepsy wearable Docking station Sunnyvale, California - 20 Mar 2019 TOWARDS AN ACUTE ONE Battery power 65mAh 65mAh 100mAh / Wireless Connectivity BLE & Proprietary BLE & Proprietary BLE & Proprietary BLE, WiFi, LTE, GSM CPU ARM M4 @ 64MHz ARM M4 @ 64MHz ARM M4 @ 64MHz ARM A53 (Quad-Core) Sensors 1 lead ECG @ 1ksps 3 DOF ACM @ 200sps 3 lead EEG @ 250sps / 3 DOF ACM @ 50sps 3 DOF ROT @ 200sps 3 channels PPG @ 250sps 3 DOF ACM @ 250 sps Memory 256MB (31 hours) 256MB (30 hours) 512MB (32 hours) 64GB Power consumption Target: 2000uA Target: 2000uA Target: 3125uA Sensors 300uA 250uA 1500uA Acquisition (RF + Logging) 1000uA 1000uA 1000uA Tiny ML 700uA 750uA 625uA /18 13

  14. EMBEDDED SEIZURE DETECTION BASED ON EEG Requirements: TRANSFORMING EPILEPSY FROM A CHRONIC CONDITION Low memory usage + low computational complexity Assumptions: Sunnyvale, California - 20 Mar 2019 Seizure spatio (+ temporal) signature is stationary TOWARDS AN ACUTE ONE Seizure spatio (+ temporal) signature is unique with regards to noise Approach: max-SNR filter = generalized eigenvalue problem /18 14

  15. EMBEDDED SEIZURE DETECTION BASED ON EEG Step 1: Algorithm design TRANSFORMING EPILEPSY FROM A CHRONIC CONDITION Threshold on RMS amplitude, calculated over 3 seconds (= minimal length of seizure) Training: find minimum RMS amplitude to detect all Seizure epochs Sunnyvale, California - 20 Mar 2019 TOWARDS AN ACUTE ONE Step 2: Define noise epochs Find epochs with RMS amplitude higher than RMS of seizure epochs. These epochs are referred to as noise. Step 3: Calculate filter coefficients Characterize Seizure and Noise covariance matrices Solve generalized eigenvalue problem /18 15

  16. EMBEDDED SEIZURE DETECTION BASED ON EEG Filter Results 16 TRANSFORMING EPILEPSY FROM A CHRONIC CONDITION /18 TOWARDS AN ACUTE ONE Sunnyvale, California - 20 Mar 2019

  17. EMBEDDED SEIZURE DETECTION BASED ON EEG Algorithm Results TRANSFORMING EPILEPSY FROM A CHRONIC CONDITION Memory : 3 seconds of RMS values (int16) + N filter coefficients (float) + N samples (int16) Complexity : N additions + N multiplications + 1 square operation (apply filter) + 1 addition + 1 subtraction (update RMS) + 1 threshold operation Sunnyvale, California - 20 Mar 2019 TOWARDS AN ACUTE ONE /18 17

  18. OUR VISION: ONLINE SEIZURE PREDICTION TRANSFORMING EPILEPSY FROM A CHRONIC CONDITION Epilepsy = a condition Defined by unpredictability Sunnyvale, California - 20 Mar 2019 TOWARDS AN ACUTE ONE 33% Unprovoked, Educational problems Recurrent seizures Limited employability Lack of control No treatment No driver license available Societal stigma ... /18 18

  19. TRANSFORMING EPILEPSY FROM A CHRONIC CONDITION Sunnyvale, California - 20 Mar 2019 TOWARDS AN ACUTE ONE tinyML Summit Transforming Epilepsy from a chronic condition towards an acute one. Hans De Clercq, Byteflies Ben Vandendriessche Jonathan Dan Tomas Fiers Lieven Billiet March 20, 2019, Sunnyvale, California /18

Recommend


More recommend