Story perspectives

Stanford's SleepFM AI Predicts 130+ Health Risks from Sleep

1/7/2026

Summary

Story summary

Stanford researchers developed SleepFM, an AI model predicting 130+ health conditions from one night's sleep data.
SleepFM was trained on polysomnography data from 65,000 participants and achieved accuracy, including a 0.89 C-index for Parkinson's and 0.85 for dementia.
The model analyzes multiple physiological signals to reveal patterns indicating future health risks.
Because the dataset covers patients with suspected sleep disorders, findings suggest sleep patterns could enable earlier interventions.

1 of 1

Story summary

Stanford researchers developed SleepFM, an AI model predicting 130+ health conditions from one night's sleep data.
SleepFM was trained on polysomnography data from 65,000 participants and achieved accuracy, including a 0.89 C-index for Parkinson's and 0.85 for dementia.
The model analyzes multiple physiological signals to reveal patterns indicating future health risks.
Because the dataset covers patients with suspected sleep disorders, findings suggest sleep patterns could enable earlier interventions.