Mamba-fusion for privacy-preserving disease prediction

Abstract

Accurate disease prediction is essential for improving patient outcomes. Privacy regulations like GDPR and HIPAA limit data sharing, hindering the development of robust predictive models across institutions. FL and multi-modal fusion frameworks counter these problems but are restricted in scalability, inter-client communication, and heterogeneity of data modalities. Techniques which provide privacy on data have an issue whereby they cause a reduction in performance or are computationally costly. This paper presents Mamba-Fusion for Disease prediction, a privacy-preserving framework for multi-modal data. It uses a hierarchical FL architecture to minimize the communication costs and improve the architecture's scalability solution and a Mixture of Experts (MoE) with LSTM based layers for dynamic temporal integration. The latest techniques like, differential privacy, secure aggregation protect both the data and its accuracy of the data as well. Experimental results on multi-modal clinical measurements, ECG, EEG, clinical notes, and demographic data support the applied framework. We have then used Mamba-Fusion to achieve 92:4% accuracy, 0:91 F-Score, and 0:96 AUC-ROC by keeping the privacy leakage at 0:02 and communication costs to 12:5 MB, which make it superior to conventional FL techniques. These results affirm Mamba-Fusion as an applications that are secure enough to support collaborative healthcare analytics on a large scale.

Keywords: Chronic disease prediction; Differential privacy in healthcare; Hierarchical aggregation mechanism; Multi-modal data fusion; Privacy-preserving federated learning; Scalable healthcare analytics.

Fig. 1

Mamba-Fusion for Privacy-Preserving Disease Prediction.

Fig. 2

Federated Learning in Healthcare.

Fig. 3

Federated Learning in Healthcare.

Algorithm 1

Federated Learning with Hierarchical Aggregation for Mamba-Fusion

Fig. 4

Performance Metrics Comparison Across Models. Mamba-Fusion consistently outperforms Centralized LSTM and Federated GRU across all key metrics. Accuracy is shown as the highest metric achieved by Mamba-Fusion (92.4%).

Fig. 5

Privacy Leakage with Communication Efficiency. The figure showcases the superior privacy-preserving properties of Mamba-Fusion compared to the baseline models. The privacy leakage rate for Mamba-Fusion is only 0.02, significantly lower than its counterparts.

Fig. 6

Communication Cost and Latency Comparison. Mamba-Fusion demonstrates the lowest communication cost and latency, making it highly efficient for real-world federated setups.

Fig. 7

Loss Convergence Over Epochs. The training and validation losses for Mamba-Fusion exhibit stable convergence, achieving the lowest final loss among the models.

Fig. 8

Performance Comparison of Mamba-Fusion and Baseline Models: The graph compares the convergence of loss over epochs for Mamba-Fusion, Centralized LSTM, and Federated GRU models. Mamba-Fusion demonstrates superior convergence with lower loss.

Fig. 9

ROC Curve Comparison Across Models. The AUC-ROC of the Mamba-Fusion is significantly high at 0.96 as indicated in the ROC curve showing the classifiers ability to classify instances between the different classes.

Fig. 10

Confusion Matrix with Additional Categories. The performance of Mamba-Fusion for three categories is presented in the matrix along with class-wise precision and recall percentages.