GShC-Net: Hybrid deep learning with DCTLAP feature extraction for brain tumor detection

Abstract

A brain tumor is an abnormal cell growth in a brain, which is not detected early. Initial detection of brain tumors is extremely critical for treatment planning as well as the survival of a patient. Brain tumors come in different forms, have unique properties, and require tailored therapies. Thus, detecting brain tumors physically is a laborious, complex, as well as error-prone process. Hence, an automated computer-assisted diagnosis with better correctness is presently in high demand. Here, this paper developed a hybrid GoogleNet-Shepard Convolutional Networks (GShC-Net) method that is employed for detecting brain tumors. The process of this approach is as illustrated follows. Firstly, an input image is carried out from the database that is given to a pre-processing module. After that, brain tumor segmentation is performed, as well as features such as Haralick texture features, Statistical features, and Discrete Cosine Transform with Local Arc Pattern (DCTLAP) are extracted. Finally, brain tumor is detected based on GShC-Net. Moreover, the GoogleNet and Shepard Convolutional Neural Networks (ShCNN) models are fused to create GShC-Net, in which the layers are modified. The proposed GShC-Net method effectively improves the early detection and classification of brain tumors, potentially aiding in more accurate and timely medical diagnoses. Furthermore, the GShC-Net is assessed by using True Positive Rate (TPR), True Negative Rate (TNR), as well as accuracy and the values attained are0.940, 0.930, and 0.932, respectively.