Digital biomarkers

Parkinson biomarkers and digital measurement

A biomarker search should specify its intended question: research classification, a disease-associated characteristic, change over time or response to an intervention. Digital biomarker publications can examine wearable sensors, movement recordings, voice characteristics and computational methods. High performance within one dataset does not by itself demonstrate equivalent performance in another clinical setting.

For a sensor or machine-learning paper, examine the collection protocol, participant count, separation of training and evaluation groups and independent validation. Check whether recordings from the same person occur in both training and test sets. Read the source methods and limitations for repeatability and performance under everyday conditions. This catalog’s topic assignment is a discovery aid rather than validation of a device or model.