Adaptive and secure multimodal federated learning for distributed cardiovascular risk stratification with interpretable clinical decision support

FedSCF: sensitivity-aware collaborative fusion for personalized federated learning in medical image classification

A PFL framework, FedSCF, which models client heterogeneity at the parameter level, including a relative perturbation-based sensitivity evaluation is designed to identify critical parameters for personalized modeling, while the remaining parameters participate in cross-client sharing.

Federated Self-Supervised Learning for Privacy-Preserving Clinical Diagnostics in Distributed Healthcare Systems

The increasing adoption of distributed digital healthcare systems has enabled collaborative artificial intelligence (AI)-based clinical diagnostics across hospitals and medical institutions. However, conventional centralized learning requires sensitive patient information to be transferred to a common server, introducing substantial concerns regarding data privacy, security, institutional governance, and regulatory compliance. Furthermore, obtaining sufficiently large and accurately annotated clinical datasets remains challenging because medical annotation requires substantial expertise and resources. To address these limitations, this study proposes a Federated Self-Supervised Learning (FSSL) framework for privacy-preserving clinical diagnostics in distributed healthcare systems. The proposed architecture combines self-supervised representation learning with federated optimization, enabling participating healthcare institutions to learn informative clinical representations from locally available unlabeled data without transferring raw patient records. Each participating client performs self-supervised pretraining followed by task-specific local optimization, while only protected model updates are communicated to the aggregation server. A privacy-aware aggregation mechanism is incorporated to reduce the exposure of institution-specific information during collaborative training. The framework further addresses heterogeneous and non-independent and identically distributed (non-IID) clinical data through adaptive client aggregation and representation alignment. Experimental evaluation demonstrates that the proposed FSSL framework achieves a diagnostic accuracy of 96.18%, sensitivity of 95.47%, specificity of 96.72%, precision of 95.83%, F1-score of 95.65%, and area under the receiver operating characteristic curve (AUC) of 0.981. Compared with conventional federated supervised learning, the proposed method provides approximately 4.6% improvement in diagnostic accuracy and 5.1% improvement in F1-score, while substantially reducing dependence on labeled clinical samples. The results indicate that self-supervised representation learning can improve the effectiveness of federated clinical diagnostics under decentralized and heterogeneous healthcare environments. The proposed framework provides a scalable foundation for privacy-conscious collaborative clinical AI while keeping patient data within the originating healthcare institution and is intended to support, rather than replace, professional clinical decision-making.

Federated Prototype Learning for Cardiovascular Diseases Recognition Under Non-Independent and Non-Identically Distributed Heart Sounds.

Regulatory-orientedDeep federated learning framework for multi-hospital medical imaging: privacy-preserving, explainable, and generalizable diagnosis

The proposed FL framework provides a privacy-preserving, explainable, and computationally efficient solution for collaborative AI in medical imaging by combining adaptive federated learning, secure privacy mechanisms, and explainable AI techniques, demonstrating strong potential for deployment in multi-hospital clinical environments.

A Feature-Augmented Analytic Federated Architecture for Early Sepsis Detection

The proliferation of Internet of Medical Things devices within the predictive healthcare paradigm necessitates robust, privacy-centric collaborative learning frameworks to detect and mitigate rapid clinical deterioration. Traditional federated learning methodologies, while attempting to preserve patient data locality, are fundamentally constrained by multi-round gradient synchronization protocols, imposing prohibitive communication latency and remaining susceptible to false negatives under extreme non-independent and identically distributed conditions. To address these challenges, this study introduces the Feature-Augmented Analytic Federated (FaFL) Architecture, which fundamentally replaces iterative gradient synchronization with a single-round closed-form computational paradigm. By instituting a proactive feature mixing mechanism via a decoupled zero-knowledge proof global buffer, the proposed framework empowers local grassroots nodes to neutralize extreme clinical heterogeneity in a single phase. The architecture employs a closed-form analytic solution combined with a trace-weighted absolute aggregation protocol to rigorously guarantee stochastic convergence and absolute cryptographic resilience without requiring recursive parameter exchanges. Extensive empirical evaluations against existing baselines under severe Dirichlet non-independent and identically distributed conditions and Byzantine poisoning attacks demonstrate that the framework fundamentally eradicates high false-negative rates in resource-constrained clinics. Consequently, the proposed architecture robustly guarantees generalization stability, substantially outperforms existing paradigms in predictive fidelity and computational efficiency, and establishes a new operational standard for mission-critical clinical networks.

A Multi-Modal Cloud Architectures for Healthcare and Predictive Clinical Analytics

The data in healthcare is highly heterogeneous and is continuously increasing from Electronic Health Records (EHRs), medical imaging, wearable Internet of Things (IoT) devices, laboratory systems, and genomic sequencing, which makes it difficult to support intelligent clinical decision making. Current cloud-based healthcare systems may lack sufficient interoperability, scalability, privacy protection, and integration with multiple modalities, due to various factors. The paper proposes a novel multi-modal cloud architecture, which combines edge computing, cloud data lakes, multimodal feature fusion, explainable artificial intelligence (XAI) and federated learning technologies to achieve secure and scalable predictive clinical analytics. The architecture uses deep learning techniques, such as Convolutional Neural Networks, Vision Transformers, Long Short-Term Memory networks, Graph Neural Networks, and multimodal Transformers, to process and analyse diverse healthcare data types. The superior disease prediction accuracy (99.21%), lower latency (42 ms), greater privacy preservation and greater clinical interpretability over existing healthcare frameworks are confirmed by the experimental evaluation performed on benchmark datasets, including MIMIC-IV, eICU, CheXpert, MIMIC-CXR, PhysioNet, BRATS, NIH Chest X-ray and UK Biobank.