Detection of Brain Tumor Using Novel Convolutional Neural Network with Magnetic Resonance Imaging

Brain tumor is a dangerous form of cancer. Hence, the early detection and recognition of brain tumors is a crucial process for saving the lives of patients. Recently, deep learning has played an important role in the diagnosis of medical imaging. Deep learning-based Convolutional Neural Network (CNN) is intensively used to detect brain tumors from Magnetic Resonance Imaging (MRI) images. Automatic detection and classification of brain tumors is the most prominent step in assessing the abnormal tissues of the brain and providing the appropriate treatment in patient recovery. In this work, for the detection of brain tumors, we proposed the 20-layered novel CNN model by using 253 MRI images from the publicly available Kaggle dataset. The dataset consists of 155 MRI images of tumorous and 98 non-tumorous images. Firstly, input images are resized and split into multiple styles (60:20:20, 70:15:15, 80:10:10, 90:05:05) to get the higher performance of our proposed model. It has been found that the proposed model achieved a higher accuracy of 96%, f1-score of 96%, recall of 96%, and precision of 97% with a splitting ratio of 80:10:10 of the dataset in 20 epochs as compared to other splitting ratios for the detection of brain tumors. The main goal of this research work is to build a less layered CNN architecture and to get the best splitting ratio of the dataset, which can detect brain tumors of MRI images without augmentation and gives better performance as compared to the previous work which has been done on the same dataset.