An AI smartwatch that detects seizures and saves lives | Rosalind Picard

Episode Summary

Executive Summary: Rosalind Picard explains how wearable sensors and AI, originally developed to measure stress and emotion, unexpectedly revealed seizure signatures that can help detect generalized tonic-clonic seizures and potentially reduce SUDEP risk. She connects epilepsy monitoring, companionship, and rapid intervention to saving lives, arguing for AI designed to summon help and support people in moments of medical crisis.

Main Topics: SUDEP and epilepsy risk (Priority: 5/5): Picard introduces SUDEP as a little-known but deadly outcome of epilepsy, using Henry’s story to show why awareness matters and why companionship and medication adherence can reduce risk. From emotion sensing to seizure detection (Priority: 5/5): Her lab’s work on stress and skin conductance led to wearable sensors that captured unexpected physiological patterns during seizures, opening a path from affective computing to neurology. Wearable data and AI/ML detection (Priority: 5/5): By combining electrodermal activity, motion, EEG, ECG, and machine learning, the team built more accurate automated seizure detection than simple shake detectors. Physiology of seizures and SUDEP timing (Priority: 4/5): Picard explains that after generalized tonic-clonic seizures, breathing can stop before the heart stops, creating a window for intervention. Clinical collaboration and commercialization (Priority: 4/5): Partnerships with hospitals and Empatica transformed research prototypes into an FDA-approved smartwatch for seizure detection and alerting. AI as a tool for human support (Priority: 5/5): The talk ends with a call to build AI that helps people summon assistance, especially for conditions people may be reluctant to disclose.

Key Arguments: SUDEP is common enough to matter but remains underrecognized; awareness is low despite its severity. Two scientifically supported ways to reduce SUDEP risk are taking prescribed medication and ensuring someone is present during a seizure. Wearable electrodermal activity can reveal strong physiological changes before or during seizures, enabling detection. Machine learning improves seizure detection when multiple sensor streams are combined, outperforming simple motion-based approaches. The post-seizure period can include breathing cessation before cardiac arrest, meaning timely human intervention can save lives. AI should be designed not just for automation, but to connect vulnerable people with nearby human help.

Data Points: SUDEP frequency: every 7 to 9 minutes - Average rate cited for sudden unexpected death in epilepsy SUDEP deaths in the U.S.: more than sudden infant death syndrome each year - Comparison used to highlight underawareness Epilepsy control with medication: two-thirds - Proportion of people with epilepsy whose condition is controlled by medication Years of potential life lost: number two among neurological disorders - SUDEP’s impact on premature mortality burden Study enrollment: 90 families - Families enrolled in the 24/7 monitoring study Seizure detection result: 100% of the first batch of grand mal seizures - Initial batch showed large skin conductance responses Age groups most affected by SUDEP: 20s, 30s, and 40s - Peak ages mentioned for SUDEP occurrence Lifetime epilepsy risk: 1 in 26 - Chance that a person will have epilepsy at some point Sensor monitoring: 24/7 - Continuous wearable data collection used in the research Signal channels: EEG, ECG, EDA, video, wrist motion - Multimodal monitoring in epilepsy monitoring units

Pivotal Quotes: "There is a SUDEP every seven to nine minutes." — Rosalind Picard: She introduces the scale of the problem and why it deserves more public attention "The second thing that reduces the risk of SUDEP is companionship." — Rosalind Picard: She emphasizes that having someone present during a seizure can be lifesaving "Let's build AI that makes everybody's lives better." — Rosalind Picard: Closing call to action about the purpose of AI and technology

Implications: The talk suggests AI and wearables can move beyond convenience into lifesaving care, especially for epilepsy. It also urges listeners to normalize disclosure, support, and rapid response so technology can connect people to human help when it matters most.