FROM TOPOLOGICAL BIOMARKERS TO CLINICAL EVIDENCE IN MENTAL HEALTH RESEARCH

Abstract

This review evaluates whether topological biomarkers in mental-health research are ready for clinical interpretation. A PRISMA-based synthesis of 38 eligible reports examined prediction, subgroup discovery, validation, and reproducibility across psychiatric conditions. Persistence-based and Mapper-derived features frequently separated patients from controls or related to symptoms and outcomes, but small samples, reused repositories, site effects, and incomplete external validation limited inference. Topological accuracy should not be equated with diagnosis. Translation requires nested model development, leave-site-out and independent testing, calibrated uncertainty, demographic reporting, and prospective links to symptoms, function, and treatment outcomes.