Brain Disease Classification from MRI Scans Using EfficientNetB0 Feature Extraction

The classification of brain diseases has proven to be a major challenge for medical professionals worldwide. Early disease detection is a necessary first step in proper treatment of brain diseases, but accurate disease detection is increasingly important due to the complexity of numerous subjective factors and ambiguities of magnetic resonance imaging (MRI) scans. In this research, a Convolutional Neural Network (CNN) model EfficientNetB0 has been proposed to identify 3 different brain diseases (Brain tumor, Alzheimer and Parkinson) from MRI scan images. The CNN model has been fine tuned for each of the disease class and the performance metrics of the disease classification has been compared for individual classification as well as combined classification. In combined classification, the proposed EfficientNetB0 model achieved a validation accuracy of 95.8 %. The result analysis indicates that the proposed CNN model has a very high classification accuracy which can assist doctors in making proper timely diagnoses.