Google is introducing a search model for continuous glucose sensors, designed to learn with little annotated clinical data.
Presented on August 26, GlucoFM learns from continuous glucose measurements by distinguishing between slow changes and short-term deviations. Its self-supervised training also stores information on missing data and the time of observations.
The study evaluates the model across four cohorts and seven clinical tasks, including diabetes risk and insulin resistance. The authors report progress in the comparative references, as well as better adaptation to new datasets. These are research findings, and do not constitute validation of a diagnostic tool that can be used without further clinical evaluation.
Read the full article on the original source
Go to the source →