CADA+ (Enhanced Cohort Adjudication Data Annotation)

Project Summary Data labeling and continuous AI model validation are essential elements in clinical AI projects. This RSE award will enable the applicant to further develop a scalable and generalizable platform to support labeling of multimodality data as well as supporting continuous model benchmarking. The design of the platform will be anchored at first satisfying needs of three active NIH projects that will need support of both structured and unstructured EHR data, MRI brain scans, and neurophysiological and cardiovascular signals. The applicant will also collaborate with external collaborators including PhysioNet and OHDSI to ensure the developed platform can be used to support a wide array of applications from research communities and produce data labels in standardized terminology. To support continuous model benchmarking, the applicant will integrate functionalities of an existing application ModelsMeetData(M2D) into the platform to publish AI model cards, hosting vested AI models, and testing them with labeled data.