Stroke Detection and Diagnosis in CT Scan Based on Deep Learning

This research introduces a system capable of automatically distinguishing between normal and stroke-affected CT scans. The model developed aims to significantly reduce the workload on radiologists by providing an efficient tool for reading and analyzing CT scans. Hence, the human error could be minimized. To develop the Deep learning-based Stroke Detection and Diagnosis (DSDD), a Bayesian Neural Network (BNN) model was employed due to its superior performance and high accuracy in medical image classification tasks. The dataset used in this research, sourced from Kaggle, comprises both normal and abnormal (stroke) CT scan images. The results of the classification demonstrate that the DSDD is highly effective, correctly classifying the majority of images within the dataset. The system achieves impressive metrics, including high accuracy, precision, recall, specificity, and an elevated F1-Score. These outcomes indicate that the DSDD is a reliable and valuable tool for the automatic detection and diagnosis of strokes in CT scan images.