Manuscript received April, 12, 2025; revised May 21, 2025; accepted June 18, 2025; published September 17, 2025.

Abstract—The proliferation of edge devices supports realtime diagnostic testing, even in rural or underserved locations. Convolutional Neural Networks (CNNs) are highly effective at analyzing medical images, including Computed Tomography (CT) scans and chest X-rays, for detecting heart diseases, but their computational complexity usually makes them unsuitable for usage on edge devices with limited resources. This paper presents a new compressed layered knowledge distillation model for precise medical image diagnosis, such as detecting COVID-19-related lung infections or identifying cardiovascular conditions. We utilize knowledge distillation to transfer the teacher network’s knowledge to a smaller, compressed student network for edge deployment. Moreover, we utilize a well-structured layer compression approach, emphasizing decoupling and merging techniques instead of pruning, to optimize the student network architecture. Two data sets, Chest CT-Scan and SARS-CoV-2 CT-Scan, were used to test the suggested model. When compared to existing models, our performance is superior. For the Chest CT-Scan dataset and SARS-CoV-2 CT-Scan, we achieved 98.93% Accuracy, 98.41% Precision, 98.69% Recall, and an F1-Score of 98.44%. The Mean Squared Error (MSE) was 0.04, with a Root Mean Squared Error (RMSE) of 0.16. For the Chest CT-Scan dataset, our results were similarly strong: 98.25% Accuracy, 98.78% Precision, 98.86% Recall, and an F1-Score of 98.14%. The MSE for this dataset was 0.09, and the RMSE was 0.13. These results verify the efficacy of our intended technique for achieving high diagnostic accuracy at low error in edge devices.

Keywords—edge computing, medical image analysis, COVID-19 diagnosis, deep learning compression, knowledge transfer, model optimization, layer fusion, efficient inference

Cite: Yogendra N. Prajapati, Dev Baloni, and Avdhesh Gupta, "Optimized Heart Disease Image Classification on Edge Devices Using Knowledge Distillation and Layer Compression," Journal of Image and Graphics, Vol. 13, No. 5, pp. 459-468, 2025.

Copyright © 2025 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC-BY-4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.