Why one biomarker is never enough in neurodegeneration.

The limit of a single measure

Alzheimer’s and Parkinson’s do not announce themselves through one channel. They change memory, movement, speech, vision and mood on overlapping timelines that differ from person to person. A test built on a single biomarker, however good, sees only one slice of that, and can be thrown off by an unrelated cause, a bad night’s sleep or simple individual variation.

What multi-modal research shows

Reviews of digital biomarkers in neurodegeneration report that combining modalities, such as movement, speech and eye or facial data, tends to give a more reliable read than any one alone (Machine learning and digital biomarkers, 2024). Newer multi-modal frameworks integrate imaging with facial features and argue that fusion helps with the generalisation problems single-modality models face (Springer 2026).

Deep Medicine’s multi-modal design

Deep Medicine’s platform is designed to fuse five biomarker streams, voice, eye tracking, facial expression, balance and neurocognition, captured together in one session and read by a single multi-modal model. The intent is that each stream adds context to the others, so the picture is harder to fool than any single test, and the model is built to improve as more sessions are added.