An Efficient Clinical Decision Support Framework Using IoMT Based on Explainable and Trustworthy Artificial Intelligence with Transformer Model and Blockchain-Integrated Chunking

Abstract

Background/Objectives: The use of edge-cloud architectures has increased rapidly to move the analysis of AI-enabled health data to global environments. However, data security, communication overhead, cost-effectiveness, and data transmission losses are still important problems to be solved. Methods: In this paper, we propose a reliable, explainable, and energy-efficient stress detection framework supported by a cost-oriented blockchain-based content-defined chunking approach to minimise the losses during data transfer. In the proposed architecture, the Nurse Stress dataset represents IoMT data. While the chunking process reduces communication volume and storage costs by avoiding data duplication, blockchain technology eliminates the risks of unauthorised access and manipulation by ensuring the immutability and traceability of data blocks. Results: All Transformer-based models have demonstrated over 99% accuracy. The TimesNet model, in particular, has been designated as the system's reference model, exhibiting superior performance in terms of both stability and accuracy. The main contribution of this study lies in proposing one of the first integrated frameworks that jointly employs chunking-based data management, blockchain-enabled trust mechanisms, and edge-cloud computing with XAI to ensure secure and transparent IoMT data processing. The proposed system not only performs highly accurate stress detection, but also optimises the dimensions of reliable data transmission, energy and cost efficiency, and clinical reliability. Conclusions: In this respect, the study presents a scalable, reliable, and repeatable approach in health decision support systems by combining data security, integrity, and explainability issues, which are addressed separately in the literature, in a holistic manner.