Enhancing Detection Accuracy of Brain Stroke Through Cropped CT Scans With CNN Architectures

A stroke is a life-threatening condition caused by a sudden disruption in the brain’s blood supply, potentially resulting in severe neurological damage or even death. Early and accurate identification of strokes is vital for prompt treatment and better health outcomes. This study utilizes advanced deep learning architectures to predict strokes and non-strokes from CT (Computed Tomography) images. Cropping techniques are employed to reduce noise and enhance image quality. Various deep learning algorithms, including EfficientNetB0, VGG16, MobileNet, and DenseNet, are explored, Notably, this technique has increased the accuracy to 98.7% which is achieved by EfficientNetB0 model, along with a strong recall of 97%, an F1 score of 94%, and precision at 95%. By improving stroke detection, this approach can lead to quicker treatment and better recovery chances for patients.