Enhancing Cardiovascular Disease Detection in Diabetic Patients Using CardioNet-DVS

This study introduces CardioNet-DVS, a hybrid model designed specifically to address the complex task of detecting cardiovascular diseases (CVD) by utilizing the combined capabilities of DenseNet, VGG16, and Support Vector Machines (SVM). The methodology commences by obtaining and preprocessing microscopic images of blood vessels from patients with diabetic cardiomyopathy (DCM). This involves using bicubic interpolation to resize the images and employing a combination of a Medical filter and Adaptive Histogram Equalization (AHE) to suppress noise and enhance contrast. The Watershed Segmentation algorithm is utilized for segmentation, which is further enhanced by a fuzzy edge detection method to precisely outline the edges of the plaque. Features are derived through the utilization of Local Binary Patterns (LBP) as well as the pretrained DenseNet and VGG16 networks. The aforementioned features are merged and inputted into a Support Vector Machine (SVM) classifier to achieve reliable classification. Comparative analysis reveals that CardioNet-DVS outperforms traditional methods like VGG16, DenseNet, and SVM in terms of accuracy, precision, and F1 scores. CardioNet-DVS achieves accuracy, precision, and F1 scores of $99.2 \%, 99.32 \%$, and $99.4 \%$ respectively. This comprehensive approach provides a notable improvement in the early detection and planning of treatment for diabetic cardiomyopathy and coronary artery plaque, thereby enhancing the process of making clinical decisions and improving patient outcomes.