Enhanced ResNet-Vision Transformer using Optimization Techniques for Brain Stroke Detection

Early recognition of brain stroke is the key to prompt clinical management and better outcomes. Accurate identification of stroke-affected tissue from normal brain tissue structures in Computed Tomography (CT) images, however, is still a difficult task because of its subtle intensity differences and complex anatomical characteristics. In this regard, this paper proposes an Enhanced Black-winged Kite Optimized ResNet–Vision Transformer Network (HBK-ResViTNet) for automated brain stroke detection. The first step in the CT processing includes detailed noise removal, image normalization, skull stripping, and contrast enhancement in order to improve the image quality and highlight regions of interest for diagnostic purposes. Then, ResNet50 is used to generate a discriminative local spatial representation, and Vision Transformer (ViT) is used to understand the long-range contextual information and achieve global feature interaction. The complementary features generated by these two networks are effectively merged to acquire a powerful feature representation for stroke identification. Additionally, the Enhanced Black-winged Kite Optimization (EBKO) algorithm is used to tune the network hyperparameters for better convergence speed and classification accuracy. Finally, the optimized feature representations are fed into fully connected layers to classify CT images into multiple classes. The proposed framework is evaluated using the RSNA Intracranial Hemorrhage Detection Dataset, a large-scale, multi-institutional brain CT dataset containing expert annotations for five major intracranial haemorrhage subtypes. The experimental outcomes show that it attains higher accuracy, precision, recall, and F1-score than traditional deep learning techniques. The results suggest that the suggested HBK-ResViTNet provides a dependable, efficient, and scalable computer-aided system for early diagnosis of brain stroke, which would be useful in clinical decision-making and management of stroke patients.