Brain tumors are very harmful for human beings, leading to even death. Early detection of brain tumors can be the possible solution to reduce the death rate and threats. Thus, this research reflects on a unique brain tumor detection and classification method through the mechanism of Deep Convolutional Neural Networks (DCNN) on 1255 Magnetic Resonance Imaging (MRI) images. The proposed research has been applied some image processing techniques like image filtering techniques on the publicly available dataset. The dataset is classified into two identical classes: with-brain and without-brain tumours. The research also applied the mechanism of data image augmentation to enlarge dataset size. After a set of pre-trained Convolutional Neural Networks (CNN) architectures, including VGG-19, ResNet50, Inception V3, and Xception, have been adopted to build the model and extract the features from each particular image. Then, we have applied the Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) on the extracted features to reduce the dimension and the components. After that, we have applied conventional machine learning models with the six classifiers to find the accuracy of the detection and classifications. To gain goal of the research, a set of experimental data have been enumerated and interpreted. The study achieved the highest accuracy of 99% while working with ResNet50 pre-trained architecture and ensemble classifiers. Thus, we assume that the proposed solution can effectively detect and classify brain tumors.
Keywords: Convolutional Neural Networks (CNN), Magnetic Resonance Imaging (MRI), Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA)
28-Feb-2025
