Deep learning-based computer-aided diagnosis tool for brain tumor classification

In the medical field, manual analysis of diseases using numerous medical imaging modalities is a challenging task. The proposed work presents a deep learning-based Computer-Aided Diagnosis (CAD) tool for the classification and localization of brain tumors. The CAD tool is implemented to classify and localize three types of brain tumors from the CA-MRI brain dataset (in all three views, i.e., axial, sagittal, coronal). The input brain MRI dataset is pre-processed and divided into a 70% training set, 15% testing set, 15% of the validation set. The deep learning-based model ineeption- ResNet-v2 is trained on an augmented training dataset. The performance of the pre-trained deep learning model is evaluated using numerous statistical parameters. The pretrained deep learning model obtained an accuracy of 98.72% on the training set, recall rate of 99.56%, and an Area Under Curve (AUC) of 1. Finally, the CAD tool is designed with the pre-trained deep learning model, and feature maps to localize the tumor area in multi-orientation testing images.