Brain Tumor Classification Using Arconv in Convolutional Neural Network Approach

Abstract Brain tumor detection and classification from medical images is a critical task in medical diagnosis, as early and accurate identification can significantly improve patient treatment and survival rates. Manual analysis of brain Magnetic Resonance Imaging (MRI) scans by radiologists is time-consuming and may lead to human error. Therefore, automated systems based on deep learning have gained significant attention for improving diagnostic accuracy. In this study, a brain tumor classification framework based on an ArConv (Adaptive Residual Convolution) integrated Convolutional Neural Network (CNN) is proposed to enhance feature extraction and classification performance. The proposed model utilizes convolutional layers combined with the ArConv module to capture complex spatial features and improve the learning capability of the network. MRI images are first preprocessed through normalization and resizing to ensure consistency in the dataset. The processed images are then fed into the CNN architecture where ArConv layers adaptively learn important tumor-related features. The model is trained and evaluated on a publicly available brain MRI dataset containing different tumor classes. Experimental results demonstrate that the proposed ArConv-CNN model achieves improved classification accuracy, precision, recall, and F1-score compared with conventional CNN approaches. The findings indicate that integrating ArConv with CNN can effectively enhance feature representation and improve brain tumor classification performance, making it a promising approach for computer-aided medical diagnosis systems. Keyword: Brain tumour, CNN, claddification, AI, Image Processing Brain Tumor Classification, MRI Image Analysis, Deep Learning, Convolutional Neural Network, Adaptive Residual Convolution (ArConv), Medical Image Processing, Computer-Aided Diagnosis