EDCNGWONet: Enhanced Dilated Convolutional Neural Networks based Gray Wolf Optimizer for Deep Vein Thrombosis Identification and Classification

Abstract

Deep Vein Thrombosis (DVT) occurs due to the formation of blood clots within the body's veins, most commonly in the legs. A serious complication associated with DVT is pulmonary embolism (PE), which arises when a segment of the clot detaches and travels to the lungs via the bloodstream. Since diagnosing DVT through clinical methods can be time-consuming, the development of computer-aided diagnostic systems can significantly enhance efficiency. This study introduces a novel Enhanced Dilated Convolutional Neural Network designed for detecting DVT in CT and MRI scans. The process begins with denoising the CT images using adaptive Wiener filtering to enhance image clarity. These enhanced images are then processed through a fuzzy-based thresholding algorithm to segment the edges effectively. Following segmentation, the proposed EDCNGWONet deep learning model is employed to extract relevant features. An Enhanced Support Vector Machine (ESVM) is then used to classify the CT images into categories such as coronary thrombosis, venous thromboembolism, and pulmonary embolism. The performance of the proposed framework is evaluated using several metrics, including precision, specificity, accuracy, recall, and F1 score. Experimental results indicate that the model achieves an accuracy of 93.63%.