The abstract is read at the publisher; the summary is JClub's.
ProgMF learns coherent multimodal source representations from incomplete neuroimaging data without imputing missing modalities, improving source-recovery errors especially when subject overlap is low.
- Why it matters: Conventional fusion methods often require complete observations, limiting multimodal neuroimaging studies where participants share only partially overlapping modalities.
- What they did: ProgMF sequentially integrates modalities using independent vector analysis, cross-stage component alignment, and anchor-guided representation transfer, evaluated on four synthetic modalities and then on ADNI MRI, resting-state connectivity, and 18F-florbetapir PET.
- The result: In ADNI, ProgMF yields higher mean cross-modal fusion scores than imputation-based baselines and a modest gain over complete-case IVA, and among 293 complete participants, two recovered sources associate with diagnosis, cognition, and dementia severity.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.