Metraipolis › archive › dc_4b33484b025e
CONCORDAT: Trustworthy multimodal federated learning for autism research under incomplete, heterogeneous and privacy-constrained health data (2026) · DOI 10.21203/rs.3.rs-10787425/v1
- Source
- Metraipolis archive
- Original
- https://doi.org/10.21203/rs.3.rs-10787425/v1
- License (as archived)
- CC-BY-4.0
- Archived
- 2026-10-08T00:06:20.219Z
- SHA-256 of the archived text
0cf89ccd10977ba6626bdbd206e16efacca0f9141a5e942753d1fc7dea6b0f83- Stable URL of this page
https://metraipolis.com/documento/dc_4b33484b025e
How to cite
CONCORDAT: Trustworthy multimodal federated learning for autism research under incomplete, heterogeneous and privacy-constrained health data (2026) · DOI 10.21203/rs.3.rs-10787425/v1 — Metraipolis archive, https://doi.org/10.21203/rs.3.rs-10787425/v1 (CC-BY-4.0); archived copy: https://metraipolis.com/documento/dc_4b33484b025e (Metraipolis archive dc_4b33484b025e, sha256 0cf89ccd10977ba6626bdbd206e16efacca0f9141a5e942753d1fc7dea6b0f83)
BibTeX
@misc{dc4b33484b025e,
title = {CONCORDAT: Trustworthy multimodal federated learning for autism research under incomplete, heterogeneous and privacy-constrained health data (2026) · DOI 10.21203/rs.3.rs-10787425/v1},
howpublished = {Metraipolis archive},
url = {https://doi.org/10.21203/rs.3.rs-10787425/v1},
note = {License CC-BY-4.0; archived copy https://metraipolis.com/documento/dc_4b33484b025e; sha256 0cf89ccd10977ba6626bdbd206e16efacca0f9141a5e942753d1fc7dea6b0f83},
year = {2026}
}Archived text
Reproduced under CC-BY-4.0. Attribution: Metraipolis archive, https://doi.org/10.21203/rs.3.rs-10787425/v1. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Articolo recente con licenza aperta: scheda dai metadati Crossref (posted-content).
Titolo: CONCORDAT: Trustworthy multimodal federated learning for autism research under incomplete, heterogeneous and privacy-constrained health data
Autori: Jarin Alam Prity
Pubblicato: 2026-08-25 · editore: Springer Science and Business Media LLC
DOI: https://doi.org/10.21203/rs.3.rs-10787425/v1
Licenza dell'articolo: CC-BY-4.0
Abstract (testo degli autori, CC-BY-4.0):
Abstract Multimodal brain-health research combines imaging-derived measurements with demographic, phenotypic and behavioural information, but these data are often incomplete, heterogeneous across sites, and subject to governance constraints that prevent subject-level pooling. CONCORDAT is a cross-silo federated framework for autism spectrum disorder (ASD) research that integrates these complementary feature blocks while keeping subject-level records at their source institution. Across the 20-site ABIDE-I cohort (n = 1035), adding demographic, phenotypic and Social Responsiveness Scale blocks to imaging-derived quality-control features improved 5-fold, site-stratified AUC by approximately 0.04 (0.610 to 0.648), and federated training reproduced the corresponding centralised results to within 0.001 AUC. We then stress-tested this multimodal pipeline under real acquisition incompleteness, comparing eight missing-modality scenarios; balanced accuracy fell despite nearly unchanged AUC when phenotypic information was unavailable, showing why multimodal deployment must monitor operating-point performance rather than discrimination alone. We further quantify site-harmonisation choices under different disclosure models, a client-level differential-privacy sweep (ε ∈ {1, 2, 5, 10, ∞}), clip-norm sensitivity, and a fairness intervention that raised female-ASD sensitivity by 0.532 (95% CI 0.419–0.661) for a 0.039 loss in balanced accuracy (95% CI 0.009–0.064). An independent structural-connectivity cohort (n = 53, d = 647) validates operation in a high-dimensional setting but yields a permutation-tested null (p = 0.965), which is reported rather than omitted. CONCORDAT therefore treats multimodal integration, privacy, missingness and subgroup equity as linked deployment constraints, providing an auditable framework for multi-site brain-health research rather than a clinical diagnostic tool.
Fonte: https://api.crossref.org/works/10.21203%2Frs.3.rs-10787425%2Fv1 — metadati Crossref, consultati il 2026-10-07. Interrogazione: «federated learning privacy».