What is it about?
This research presents an explainable TinyML-based framework for automated pneumonia diagnosis using chest X-ray images. The framework combines lightweight deep learning models, post-training quantization, and Explainable AI (XAI) techniques to enable accurate and transparent pneumonia detection in resource-constrained environments. Multiple deep learning architectures were evaluated, and a quantized MobileNetV3 model was selected for its balance between diagnostic accuracy, computational efficiency, and deployment feasibility. The system further integrates LIME, Shapley Value Sampling, and Occlusion methods to provide visual explanations of model predictions, helping healthcare professionals understand and validate diagnostic decisions. The proposed solution aims to support faster, more accessible, and trustworthy AI-driven healthcare services.
Featured Image
Photo by CDC on Unsplash
Why is it important?
Pneumonia remains one of the leading causes of illness and mortality worldwide, particularly in regions with limited medical resources. While deep learning has demonstrated high diagnostic accuracy, many existing models are computationally expensive and operate as "black boxes," limiting clinical trust and adoption. This work addresses both challenges by combining TinyML for efficient edge deployment and Explainable AI for transparency. The framework enables rapid and accurate diagnosis on low-resource devices while providing interpretable explanations for predictions, making AI-assisted healthcare more accessible, reliable, and suitable for real-world clinical environments.
Perspectives
This work contributes to the advancement of Healthcare 5.0 by demonstrating how efficient and explainable AI systems can be integrated into medical diagnosis workflows. From a technological perspective, it highlights the potential of TinyML to bring advanced healthcare solutions to edge devices with limited computational resources. From a clinical perspective, the inclusion of XAI strengthens trust, accountability, and collaboration between AI systems and healthcare professionals. Future developments may include larger and more diverse datasets, continual learning capabilities, transformer-based architectures, and multi-disease diagnostic support, enabling broader deployment of transparent AI solutions across digital health ecosystems.
Deep Joshi
Nirma University of Science and Technology
Read the Original
This page is a summary of: Illuminating Pneumonia: Explainable TinyML-Based Pneumonia Diagnosis Framework for Digital Health, April 2026, Institute of Electrical & Electronics Engineers (IEEE),
DOI: 10.1109/icwr69602.2026.11513312.
You can read the full text:
Contributors
The following have contributed to this page
