Detecting Hypertrophic Cardiomyopathy: A Deep Learning Approach with CNNs and Swin Transformers

Hypertrophic cardiomyopathy (HCM) is a serious cardiac illness for which the purposeful diagnosis is necessary for timely therapy.A new block based convolutional neural network (CNN) architecture specifically designed for the purpose of enhancing the accuracy of diagnosis of cardiac magnetic resonance (CMR) imaging.It takes advantage of a special block structure for HCM diagnosis that achieves superior performance to accuracy trade off over conventional approaches.This approach is a key innovation in which the fractional calculus technique, namely the Caputo derivative, is integrated into the deep learning framework.The inclusion allows the feature representation to be this enhanced, leading to much higher classification accuracy and precision.Fractional calculus in deep learning is an important contribution towards development of medical image analysis.Additionally, we compared in detail with Swin Transformers and found that although they showed good performance in other fields, they did not perform well for HCM diagnosis based on CMR.The finding confirms the limited capacity of Swin Transformers for this task, and the proposed CNN based approach remains effective.Data preprocessing techniques such as resizing, normalization and data augmentation has been used to improve the generalization of model.The metrics of high precision (93.47%), recall, F1 score and were 92.46% accuracy on the "danialsharifrazi/hypertrophic cardiomyopathy-dataset."These results suggest more effective diagnostic accuracy in HCM via cardiac CMR imaging by combining deep learning with the use of advanced mathematical techniques, such as the Caputo derivative.