Quick diagnosis and treatment are essential for the best possible outcome in cases of brain stroke, a serious medical condition characterized by an abrupt stoppage of blood supply to the brain, which can result in tissue damage and even death. The diagnosis and evaluation of brain stroke severity are greatly aided by Computed Tomography (CT) scans, which provide comprehensive information on the structure of the brain and can be used to identify stroke indications such as bleeding, blockages, or ischemic lesions. In a deep learning approach, we propose to use Convolutional Neural Networks (CNNs) to improve the identification of brain stroke in CT scan pictures. Our model's primary architecture for feature extraction is EfficientNet B4, which is renowned for its ability to successfully extract useful features from images of medical conditions. By integrating EfficientNet B4, we hope to improve brain stroke detection accuracy while maintaining computing efficiency. Our deep learning method uses the most recent CNN architectures along with transfer learning techniques to target an accuracy rate of 98%. By using this technique, medical professionals should be able to diagnose brain strokes more quickly, which should lead to better patient outcomes in clinical settings.