Manuscript received January 30, 2026; revised April 21, 2026; accepted June 2, 2026; published November 13, 2026.

Abstract—Analyzing Electrocardiogram (ECG) signals is essential for identifying heart arrhythmias and ensuring prompt medical intervention. However, deep learning models have shown promise in increasing ECG classification accuracy. The problems associated with little annotated data continue. This research investigates how the GoogLeNet network model handles this problem with transfer learning to obtain high accuracy in ECG classification. Using transfer learning from a sizable dataset, this study used a deep learning framework to optimize the GoogLeNet architecture on a collection of ECG signals and classified them into 4 classes. This study trains the dataset from the Massachusetts Institute of Technology-Beth Israel Hospital (MIT-BIH) Arrhythmia Database. This paper involved pre-processing, data enhancement, and hyperparameter optimization to enhance model performance. The results of the categorization were extensively assessed, and the methods were compared. This study achieved an incredible classification rate of 99.17% in the detection of ECG arrhythmias. The implementation of the GoogLeNet network model based on the transfer learning approach to classifying ECGs has significantly enhanced the monitoring of cardiac health.

Keywords—Electrocardiogram (ECG) classification, GoogLeNet, transfer learning, deep learning, cardiac arrhythmia, confusion matrix

Cite: Haseeb Ahmad, Ghulam Abbas, Muhammad Uzair, Nathirulla Sherif, and Zohaib Mushtaq, "Enhancing ECG Classification Accuracy by Deep Transfer Learning with Resource-limited Settings," Journal of Image and Graphics, Vol. 14, No. 6, pp. 954-965, 2026.

Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).