Development of Hybrid Deep Learning Network for Precise Segmentation of Cardiac Disease with Adaptive Nested Dilated Residual Densenet-Based Classification

Cardiovascular diseases (CVDs) remain the primary reason for mortality globally, impacting millions of individuals. Medical research and therapies depend on the timely and accurate identification of heart diseases. Individuals can be affected by different kinds of cardiac disease according to factors like cholesterol, blood pressure, and cardiac activity. Diagnostic techniques include blood analysis, electrocardiograms (ECG), and imaging techniques that remain a significant approach to identifying CVD. Cardiac magnetic resonance imaging (CMRI) has become more common for identifying and evaluating illness, scheduling medication, and detecting CVDs. Therefore, an advanced cardiac disease classification model using deep learning is developed to increase the robustness of cardiac disease diagnosis by addressing key challenges in segmentation and classification procedures. At first, the CMRI images are accumulated from standard sources. The aggregated image is segmented using the developed transformer-based hybrid network (Trans-HyNet). This was created by combining a fully convergent network (FCN) with the DenseNet technique. The partitioned images are sent to the categorization process performed via the developed adaptive nested dilated residual densenet (AND-RDNet). The ability of the classification technique is enhanced by tuning the parameters via modified position updating carpet weaver optimization (MPU-CWO). Precise localization of cardiac structures helps to detect subtle abnormalities in input images. Heart failure and other heart-related diseases are cured by accurately predicting the cardiac disease. At last, the robustness of the developed method is examined by comparing the outcomes with existing techniques. The developed method attains a superior performance of 96% accuracy, 94% sensitivity, 97% specificity, and 96% precision in terms of the Tanh activation function.