Adapting Machine Learning Diagnostic Models to New Populations Using a Small Amount of Data: Results from Clinical Neuroscience.

ABSTRACT: Machine learning (ML) is revolutionizing many areas of engineering and science, including healthcare. However, it is also facing a reproducibility crisis, especially in healthcare. ML models that are carefully constructed from and evaluated on data from one part of the population may not generalize well on data from a different population group, or acquisition instrument settings and acquisition protocols. We tackle this problem in the context of neuroimaging of Alzheimer's disease (AD), schizophrenia (SZ) and brain aging. We develop a weighted empirical risk minimization approach that optimally combines data from a source group, e.g., subjects are stratified by attributes such as sex, age group, race and clinical cohort to make predictions on a target group, e.g., other sex, age group, et

SUBMITTER: Wang R 

PROVIDER: S-EPMC11419182 | biostudies-literature | 2024 Sep

REPOSITORIES: biostudies-literature