Accurate Brain Tumor and Lobes Symmetry Detection Using Deep CNNs

Lobar asymmetries in brain MRI images can indicate neurological disorders, including tumors or structural abnormalities. Manual identification of these asymmetries by specialists is time-consuming and prone to human error, especially when handling large volumes of data. Therefore, developing automated deep learning-based approaches can improve both the accuracy and efficiency of detection. In this study, we propose a method for detecting symmetric and asymmetric brain lobes using the publicly available Brain Symmetry Lobes Dataset. MRI images were preprocessed by resizing to ۳۲×۳۲ pixels and normalizing pixel values. The processed images were then fed into a convolutional neural network (CNN) consisting of four convolutional layers with batch normalization, max-pooling, L۲ regularization, and dropout. The model was trained for ۱۵ epochs using the AdamW optimizer. Experimental results demonstrated that the proposed model achieved an overall accuracy of %۹۸.۹۰, with weighted average Precision, Recall, and F۱-Score of %۹۸.۹۱, %۹۸.۹۰, and %۹۸.۹۰, respectively, showing effective discrimination between symmetric and asymmetric lobes. ROC analysis and AUC values further confirmed the model’s robust performance in multiclass classification. These results suggest that the proposed CNN-based approach can serve as an efficient tool to assist radiologists in the rapid and accurate detection of brain lobe asymmetries

Keywords:

Brain MRI , Lobar asymmetry , Convolutional Neural Network (CNN) , Medical Image Analysis , Automated Diagnosis