Research & academia
GlucoFM: Foundation model for continuous glucose monitoring
GlucoFM separates slow glycemic trends from short-term CGM deviations, then tests whether those learned representations help with clinical prediction when labeled data are scarce.
Why it matters: The results suggest that representations learned from unlabeled CGM could make limited clinical labels go further, including in transfer and few-shot settings. They are evaluations reported by the research team, not evidence of prospective clinical utility.