EUVIMEDEUVIMED entdecken
CrossrefInfokarte

APB‐FLDPA: Adaptive Personalized Blockchain‐Federated Learning With Differential Privacy and Attention for Privacy‐Preserving Healthcare Analytics

Md Kamran Hussin Chowdhury, Proloy Kumar Mondal, Md Ariful Islam Mozumder, Hee Cheol Kim, Haewon Byeon

Healthcare Technology Letters · 2026-01 · 13 (1)

DOI: 10.1049/htl2.70079 ↗
AbstractBibliografische Angaben

Abstract

ABSTRACT Developing robust medical artificial intelligence (AI) requires collaboration across multiple institutions, but strict data protection regulations such as HIPAA and GDPR prevent centralized patient data sharing. Existing federated learning (FL) methods often exhibit 15%–30% performance degradation in real‐world clinical settings due to data heterogeneity, security threats, and privacy constraints. We present APB‐FLDPA, a privacy‐preserving federated learning framework for secure multi‐hospital disease prediction. APB‐FLDPA integrates five key innovations: (i) adaptive Byzantine‐resilient aggregation using dynamic client trust scoring, (ii) self‐attention for automated clinical feature importance, (iii) selective differential privacy applied at the final aggregation stage, (iv) cluster‐aware personalization to handle cross‐institutional heterogeneity, and (v) a lightweight blockchain module to ensure model integrity. Evaluated across five institutions using large‐scale Diabetes (183,000 patients) and Thyroid (6840 patients) datasets, APB‐FLDPA achieved 90.8% accuracy for diabetes and 83.8% accuracy for thyroid disease, with minimal performance loss (

Wissen, das zu Ihrer Arbeit passt.

Vom kostenlosen Einstieg über die persönliche Recherche bis zur vollständigen Research Suite.

Tarife ansehen →

Sind alle Verlagsvolltexte enthalten?

Nein. Bibliografische Nachweise, Abstracts und Volltexte sind unterschiedliche Inhalte. Ob ein Volltext verfügbar ist, hängt von der Quelle, Open Access und Ihren vorhandenen Lizenzen ab. Ein EUVIMED-Tarif ersetzt keine zusätzlichen Verlagsrechte.

Tarife ansehen →