COMPARATIVE EFFICIENCY ANALYSIS OF CNN AND VIT MODELS IN BRAIN CANCER DETECTION BASED ON MRI IMAGES

In this study, the task of classifying brain tumors based on magnetic resonance imaging was analyzed using deep learning approaches. The study compared convolutional neural networks and transformer-based models, including the Vision Transformer and DEiT architectures. The experiments were conducted on an open "Brain Tumor MRI Dataset" dataset, and the models were evaluated based on accuracy, precision, recall, and F1-score metrics. According to the results, the DEiT model demonstrated the highest efficiency, achieving 98.04% validation and 94.44% testing accuracy. The results obtained showed the superiority of transformer models due to their ability to effectively study global properties and are of great importance for the development of automated diagnostic systems based on MRI images.