The Rise of Invisible AI-Powered Health Tech!: Prof. Dina Katabi

Invisible Health Technology (Beyond Wearables)
📌 The core problem addressed is monitoring the health of the old and sick who often do not wear wearables or stop using them when ill.
💡 The solution is an "Emerald device," similar to a Wi-Fi box, that analyzes wireless signals bouncing off bodies to extract physiological data.
📡 This invisible technology can monitor breathing, heartbeat, sleep stages (including REM), motion, and even EEG (brain waves) without physical contact.

Capabilities and Accuracy
🤖 The system utilizes AI and custom neural networks to extract biomarkers for diseases and drug effects from reflected wireless signals.
🚶 Real-time tracking demonstrated the device accurately monitoring body movement, rendering a skeletal stick figure representation without wearables.
😴 Sleep staging accuracy achieved through this non-contact method is demonstrated to be more accurate than current wearables for sleep monitoring compared to gold standards.
🧠 For EEG monitoring during sleep, particularly in an 83-year-old Alzheimer's patient, the AI-generated signal showed high accuracy, repeating results across 20,000 individuals with a mean error less than 7%.

Disease and Drug Monitoring Applications
🅿️ Parkinson's disease detection showed ~90% accuracy in determining diagnosis based purely on respiratory signals during sleep from the wireless device, a level of insight unavailable to medical experts by simply observing breathing.
🔬 Initial predictive analysis on a longitudinal dataset showed machine learning models distinguishing individuals who later developed Parkinson's disease years later, suggesting potential for early screening.
💊 The system can detect the effect of anti-depressants from respiratory signals, achieving 84% accuracy from a single night's wireless data, helping monitor patient adherence.

Future Vision and Foundational Models
🔮 The research aims to develop a foundational model for human physiology, similar to large language models (LLMs) for text, capable of understanding various physiological signals independently.
🏥 The long-term vision is to deploy the Emerald device in every home with a chronic disease patient to transition healthcare from a reactive to a proactive system.
⚙️ This proactive monitoring will continuously extract information about diseases and therapies, allowing AI to inform patients and doctors to personalize healthcare and improve outcomes.

Key Points & Insights
➡️ The technology focuses on monitoring those who cannot or will not use traditional wearables, addressing a critical gap in healthcare for the elderly and sick.
➡️ The system’s ability to derive complex physiological data (like EEG and sleep stages) non-invasively pushes beyond current medical knowledge, as it detects insights doctors cannot currently observe (e.g., Parkinson's from breath alone).
➡️ Early promising results suggest the system can potentially predict the onset of chronic diseases like Parkinson's years before formal diagnosis based on subtle respiratory changes during sleep.
➡️ Researchers emphasized the device operates at 1,000 times lower power than standard Wi-Fi and uses Ultra Wideband technology, complying with regulatory standards.

📸 Video summarized with SummaryTube.com on Dec 10, 2025, 07:07 UTC