Attention on Personalized Clinical Decision Support System: Federated Learning Approach

TL;DR

This paper tackles privacy and data-heterogeneity challenges in building AI-driven clinical decision support by proposing a privacy-preserving federated learning framework with edge AI. It employs attention-enabled sequence-to-sequence models (Bahdanau attention) deployed on local devices, with a central server aggregating updates to form a global model that benefits from diverse, non-IID data without exposing patient information. The study demonstrates a practical implementation with five clients, using a GRU-based encoder–decoder and FedAvg-style aggregation, showing superior generalization over centralized training on heterogeneous data and an evolvable design that accommodates new symptoms and diseases without re-training from scratch. The work holds practical significance for scalable, privacy-conscious clinical data mining and personalized decision support in smart-city healthcare, while future efforts will focus on reducing communication costs and robustly handling noisy biomedical inputs.

Abstract

Health management has become a primary problem as new kinds of diseases and complex symptoms are introduced to a rapidly growing modern society. Building a better and smarter healthcare infrastructure is one of the ultimate goals of a smart city. To the best of our knowledge, neural network models are already employed to assist healthcare professionals in achieving this goal. Typically, training a neural network requires a rich amount of data but heterogeneous and vulnerable properties of clinical data introduce a challenge for the traditional centralized network. Moreover, adding new inputs to a medical database requires re-training an existing model from scratch. To tackle these challenges, we proposed a deep learning-based clinical decision support system trained and managed under a federated learning paradigm. We focused on a novel strategy to guarantee the safety of patient privacy and overcome the risk of cyberattacks while enabling large-scale clinical data mining. As a result, we can leverage rich clinical data for training each local neural network without the need for exchanging the confidential data of patients. Moreover, we implemented the proposed scheme as a sequence-to-sequence model architecture integrating the attention mechanism. Thus, our objective is to provide a personalized clinical decision support system with evolvable characteristics that can deliver accurate solutions and assist healthcare professionals in medical diagnosing.

Figures (8)

• Figure 1: System overview.
• Figure 2: Attention network architecture.
• Figure 3: Training history for centralized models.
• Figure 4: Training history for global model.
• Figure 5: Prediction of tuberculosis.