Detection of Brain Tumor Abnormality from MRI FLAIR Images using Machine Learning Techniques: Fid-Bau Portal

Detection of Brain Tumor Abnormality from MRI FLAIR Images using Machine Learning Techniques

Aswathy, A. L. / Vinod Chandra, S. S.

Brain tumor is one of the abnormalities that can affect people of all ages and result in death. Because the brain controls all other body parts, it is essential to spot any abnormalities in the brain. Magnetic resonance imaging (MRI) employs magnetic fields rather than radiation to produce images of organs. So it is used to diagnose the brain. The radiologist received a number of image slices from the MRI machines with various modalities such as T1, T1C, T2, and Fluid Attenuation Inversion Recovery (FLAIR). In brain tumors, the abnormality, that is, tumor, edema, or tumor with edema, is present only in some slices based on the tumor growth. To carry out the treatment plan, the edema or tumor present in those slices must be appropriately diagnosed. The tumor shows resemblance to the different tissues present in the brain. This makes detecting abnormalities in the brain extremely challenging. The suggested study uses the FLAIR modality to detect the abnormality since the abnormality is more evident in this modality. Here, we proposed a computer-aided system to detect the brain’s abnormalities as part of the tumor. This research also examines the effectiveness of several pre-trained networks and classifiers in detecting normal and abnormal MRI brain images. For feature extraction, it has been investigated the AlexNet, ResNet-50, ResNet-101, VGG-16, VGG-19, DenseNet-201, and Inception-v3. The evaluation was done on the performance of the Naive Bayes, Tree-based method, Discriminant method, Ensemble-based method, K-nearest neighbor (KNN), and support vector machine algorithms for classification. By analyzing, it is observed that the AlexNet with KNN gives an accuracy of 97.79%.