Real-time Brain Stroke Detection using CT Scans

Brain stroke is a serious medical condition that requires timely and accurate diagnosis to save lives and reduce long-term complications. In this study, we explore how deep learning can be applied to classify brain CT scans into two categories: stroke and normal. We evaluated four cutting edge convolutional neural network architectures—DenseNetl21, VGG16, VGG19, and ResNet50 and found DenseNetl21 to be the most effective, achieving an impressive test accuracy of 98.63%. To notice the challenge of class imbalance in the dataset, we used data augmentation approaches, which improved model’s training and generalization capabilities. These findings underscore DenseNetl21’s potential to deliver precise and efficient stroke detection, paving the way for its use in real-time clinical settings. Moving forward, our research will focus on localizing stroke affected regions, classifying different stroke subtypes, and implementing the model in diverse clinical environments to enhance diagnostic workflows.