CNN-Based Approach for the Classification of Hypertrophic Cardiomyopathy

Abstract

Cardiovascular Magnetic Resonance (CMR) imaging plays a crucial role in diagnosing heart conditions, including cardiomyopathies. With the increasing use of deep learning in medical image analysis, these techniques have shown promising results in automatically interpreting CMR images, especially for conditions like hypertrophic cardiomyopathy (HCM). This paper addresses the need for accessible and efficient diagnostic tools for CMR imaging and proposes a deep learning-based approach to classify HCM images. The methodology includes preprocessing steps such as resizing, normalization, and data augmentation to improve the model's generalization. A novel CNN-block-based model and an EfficientNetV2S-based model were developed for the classification of hypertrophic cardiomyopathy (HCM) images. Both models were evaluated on a publicly accessible collection of hypertrophic cardiomyopathy images. The results indicate significant improvements in diagnostic accuracy and other key metrics. The CNN-based model achieved an accuracy of 91.51%, precision of 90.24%, recall of 86.36%, F1 score of 88.26%, AUC of 90.44%, and MCC of 81.66%. The EfficientNetV2S model demonstrated even better performance, with an accuracy of 98.61%, precision of 98.43%, recall of 97.80%, F1 score of 98.11%, AUC of 98.44%, and MCC of 97.02%.