Stanford AI analyzes sleep data to predict more than 100 serious diseases

STANFORD, Calif. — Scientists at Stanford Medicine have developed an artificial intelligence system capable of estimating a person’s risk for more than 100 different diseases by analyzing biological signals from a single night of sleep.

The system, called SleepFM, was trained using nearly 600,000 hours of sleep recordings collected from 65,000 individuals.

Researchers utilized data from polysomnography, a comprehensive clinical test that tracks brain activity, heart function, breathing patterns, and other physical signals. While this method is the gold standard for diagnosing sleep disorders, the study found it also records a massive amount of untapped physiological data.

SleepFM is built as a "foundation model," similar to the technology behind ChatGPT, but it is trained on biological signals instead of text.

Following its training, the model demonstrated a high level of accuracy in predicting future illnesses. The strongest results were recorded for cancers, pregnancy complications, circulatory diseases, and mental health disorders.

Specifically, SleepFM showed high precision in predicting Parkinson’s disease, dementia, and prostate and breast cancers.

The significance of SleepFM lies not merely in the application of artificial intelligence to clinical medicine, but in its fundamental shift in methodology. By leveraging a "foundation model" architecture, researchers have developed a system capable of deciphering the "language of sleep" through complex, multimodal biological data streams. This approach offers a far more efficient alternative to the fragmented, task-specific models that currently dominate the field.

According to an analysis by the Saigon Sentinel, the breakthrough’s primary value is its ability to unlock "dark data"—vast repositories of medical information collected during routine clinical practice that have historically remained under-analyzed. For decades, the predictive potential of polysomnography records has been largely ignored. SleepFM demonstrates that these routine diagnostic tests may harbor latent biomarkers for diseases that manifest years in the future.

This pivot from reactive diagnosis to long-term prognosis marks a critical evolution in preventive care. Rather than merely identifying immediate conditions such as obstructive sleep apnea, the technology positions sleep studies as a comprehensive health screening tool capable of providing early warnings for cancer, cardiovascular disease, and cognitive decline.

However, the path to clinical integration faces significant hurdles. The model requires rigorous external validation across more diverse populations to move beyond its initial single-center data source. Furthermore, policymakers and providers must navigate the ethical and logistical complexities of communicating sensitive risk predictions to patients. Until the high costs and limited accessibility of specialized sleep testing are addressed, the widespread adoption of such advanced diagnostic tools remains a significant challenge for health equity.

While this technology is currently in the research phase and has yet to make a direct impact, its long-term potential for the Vietnamese-American community is significant. Once it becomes widely accessible and affordable, it could serve as a vital health screening tool for our families. From entrepreneurs in the nail salon industry to those running phở restaurants across various Little Saigons, this innovation would allow for the early detection of serious health risks through non-invasive means, offering a simpler alternative to complex or intimidating medical procedures.