Leveraging Fuzzy C-Means and EfficientNetB0 for Improved ECG Image Classification

Electrocardiogram (ECG) diagnosis techniques provide more precise and more straightforward access to cardiovascular disease (CVD) detection than current clinical procedures allow. The presented framework combines EfficientNetB0 with Fuzzy C-Means (FCM) clustering to establish an innovative deep-learning hybrid model that optimizes ECG image classification. The system initiates its workflow with complete image processing that includes dimension adjustments along with normalization steps and color transformation before using EfficientNetB0 for deep feature acquisition. Using FCM clustering provides this method with a unique capability to generate fuzzy membership values that capture unpredictability and deviation found within ECG data. Two benchmark datasets undergo 5-fold cross-validation testing for evaluation of the model, with the first containing 1,376 ECG images spread across four diagnostic classes and the second having 3,264 images. Experimental tests show that EfficientNetB0-FCM produces better classification accuracy at 99.57% than other current CNN methods depending only on raw image features. Fuzzy membership elements are assessed regarding their ability to improve both precision and generalization using analyses that demonstrate the effectiveness of combined clustering.