A Comparative Analysis and Visualizable of MRI Image Type-Based Brain Tumor Classification Using Transfer Learning Models

A brain tumor is an abnormal growth of cells in the brain that damage normal brain function. Now, it is the deadliest illness, brain tumors have a very low prognosis when they reach their greatest degree. Misdiagnosis of brain tumors will lead to incorrect medical intervention and lower patient survival rates. A precise identification of brain tumors is essential for developing an effective treatment strategy that will prolong the lives of individuals with the condition. In our proposed work, we choose the “Brain Tumor MRI Images 44 Classes” dataset. We preprocessed the dataset using five transfer learning models using the image size
$\mathbf{224} \times \mathbf{224}$
and a total of 100 epochs for all models and those models are VGG16, VGG19, ResNet50, ResNet101, and EffiecientNetB5 in our in-research work. From all those models we got the maximum accuracy of 94.75% from our proposed EffiecientNetB5 model within 28 min for 100 epochs and 16–17 sec per epoch and we chose 70% of training and 15% of the test and also 15% validation data for all models. Therefore, expanding the dataset to include a greater variety of patient demographics and enhancing transfer learning algorithms to increase prediction accuracy will be the main goals of future research.