Masked-autoencoder fusion for ASD classification with incomplete multimodal data
Source check: Source checked. Not independent scientific confirmation.
Finding: The MMR-MF3 framework combines fMRI connectivity and behavioral features using modality-specific encoders, masked-autoencoder imputation, cross-modal transformer fusion and sex-stratified classification heads. On ABIDE I/II data, the paper reports AUC 0.89 for males and 0.86 for females and evaluates generalization with leave-one-site-out cross-validation. This is retrospective dataset evaluation; it does not establish diagnostic benefit or clinical deployment.
Technology: A masked autoencoder reconstructs missing modalities and a cross-modal transformer fuses fMRI connectivity and behavioral features for sex-stratified ASD classification.
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- A masked autoencoder reconstructs missing modalities and a cross-modal transformer fuses fMRI connectivity and behavioral features for sex-stratified ASD classification.
- Test and result
- Retrospective ABIDE I/II datasets; fivefold and leave-one-site-out validation, modality-missingness experiments, ablations and SHAP analysis.
- Limitation / still unknown
- No prospective clinical evaluation or clinical benefit was established; ABIDE retrospective validation does not establish real-world diagnostic utility.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13750264/
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.
Review date: 22 Nov 2026. Review scope: Primary full text and PubMed bibliographic record, including Methods, Results, author affiliations and first publication date..
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.