Clinical research often depends on assessments collected with different instruments at different sites or visits. Those records can contain overlapping but not identical measurement sets, making comparisons difficult. A new preprint introduces BFHD (Bidirectional Feature Harmonization Decomposition), a method designed to harmonize such heterogeneous assessments at the feature level.

According to the abstract, the motivation is to move beyond harmonizing only total scores. BFHD treats the measurements as mixed sets that need bidirectional harmonization and decomposition, though the abstract does not provide experimental details. The approach is presented as a way to align related but non-identical clinical measurement systems across centers and time.