What do 7.5 billion human brains need to thrive in the Digital Age, and what ...
Die SlideShare-Präsentation wird heruntergeladen.
×
Hier ansehen
Empfohlen
Weitere Verwandte Inhalte
Ähnlich wie Justin T. Baker, Into the Deep (20)
Weitere von The Petrie-Flom Center for Health Law Policy, Biotechnology, and Bioethics (20)
Aktuellste (20)
Justin T. Baker, Into the Deep
- 1. Into the Deep How New Tools for Studying Human Behavior Will Change Mental Health and Society (for the Better) Justin T. Baker MD, PhD Assistant Professor of Psychiatry, Harvard Medical School Scientific Director, Institute for Technology in Psychiatry, McLean Hospital HLS Computational Justice October 23, 2019
- 2. Designing Human-Centered Tools for a Digital Age
- 3. Syndromes Circuits Brain Structures Brain Systems Signs & Symptoms Genes Linking Brain and Behavioral Variation to Understand the Brain and Disease
- 4. Sensor Networks Feature Extraction Feedback Latent Construct Models Motors & Actuators
- 5. Sensor Networks Feature Extraction Feedback Latent Construct Models Motors & Actuators •
- 6. Sensor Networks Feature Extraction Feedback Latent Construct Models Motors & Actuators •
- 7. Sensor Networks Feature Extraction Feedback Latent Construct Models Motors & Actuators •
- 8. Psychiatric syndromes are heterogeneous
- 9. Schizophrenia Bipolar Disorder Schizoaffective disorder Major Depressive disorder Psychiatric syndromes are heterogeneous
- 10. Syndromes change over time
- 11. • Training can take months (research) to years (clinical) • Administration and scoring is burdensome. • Highly Subjective • Poor inter-temporal, inter-rater reliability Behavioral assessment is human resource intensive and expensive
- 12. What brought you into the hospital? Has anything in particular been on your mind recently? What are you goals for this hospitalization? How are people treating you here? How’s the food been? How’s your mood? How’s your thinking? How’s your energy? How many hours have you slept? How’s your self-confidence now compared to usual for you? Q1 Q2 Q5 Q6 Q7 Q8 Q9 Q10 Q11 Q12 N/A Facial Action Amplitude High Low ParticipantEvaluator Q1 Q2 Q3 Q6-8 Q9-10 Q10-11 Q12 end smile smile speech speech squint squint frown frown brow furrow brow raise brow raise 5-10 min brief encounter MDAlone
- 13. Unusual Thought Content (PANSS) Q2 r=-0.744 r=0.752 Depression (BPRS) r=0.656 Alone Alone Q12 Q12 r=0.739 r=-0.578 Delusions (PANSS) Negative Sxs (PANSS) brow furrow brow raise smile brow raise smile r=0.680 Negative Symptoms (PANSS) Predicted Observed What brought you into the hospital? Has anything in particular been on your mind recently? What are you goals for this hospitalization? How are people treating you here? How’s the food been? How’s your mood? How’s your thinking? How’s your energy? How many hours have you slept? How’s your self-confidence now compared to usual for you? Q1 Q2 Q5 Q6 Q7 Q8 Q9 Q10 Q11 Q12
- 14. Voice as a psychiatric VS? o ia
- 15. Voice as a psychiatric VS? PANSS Negative Total PANSS Suspiciousness PANSS Neg Blunted Affected PANSS Neg Emotional Withdrawal PANSS Neg Social Withdrawal PANSS Neg Speech Flow PANSS Gen Depression BPRS Depression BPRS Bizarre Behavior BPRS Blunted Affect BPRS Emotional Withdrawal Constricted Vowel Space Expanded Vowel SpaceTypical Vowel Space +0.50.0-0.5 o ia
- 16. Relationship of Automatic and Conventionally Acquired Measures
- 17. Leveraging technology to build robust systems for collecting real world evidence at level of the individual. “Deep Phenotyping” ~200 days
- 18. Significant Locations Two years of Geospatial Recording N/A Location classes (most to least common)
- 19. N/A Location classes (most to least common) Significant Locations Two years of Geospatial Recording
- 20. N/A Location classes (most to least common) Any Given Sunday Two years of Geospatial Recording
- 21. N/A Location classes (most to least common) Any Given Sunday Two years of Geospatial Recording •
- 22. N/A Location classes (most to least common) Any Given Sunday Two years of Geospatial Recording •
- 23. N/A Location classes (most to least common)
- 24. Latent Psychiatric Variables Feature Detection Raw Data Capture Sleep / Circadian Executive Function Language Affect Regulation Stress Sensitivity Comprehensive Phenotyping Machine Learning & Predictive Analytics Design & Consumer Focus Tools for Research & Clinical Practice The Future of Psychiatry?
- 25. Will AIs eventually replace psychiatrists? British Journal of Psychiatry Debate, 2019 (In Press)
- 26. “We will never prevent the violent autonomy of human behavior.” jtbaker@partners.org @JustinBakerMD
