Brain Tumor Grading in Histopathology Images Using MobileNetV2: A Comparative Study with ResNet50 and InceptionV3

Brain cancers classification is vital for early diagnosis and treatment planning, having a significant impact on patient survival. Manual classification is time-consuming and subject to human error, requiring deep learning-based automated classification. In this paper, a deep learning model based on the MobileNetV2 architecture, pre-trained on ImageNet, is utilized to classify Gliomas— Glioblastoma Multiforme (GBM) and LowGrade Glioma (LGG)—from histopathology images. The model is strengthened with the addition of global average pooling, batch normalization, L2-regularized dense layers, and dropout to enhance generalization. It is trained with the Adam optimizer with a learning rate scheduler to adaptively improve performance. Experimental results show that the proposed model achieves 99% accuracy, surpassing conventional CNN-based methods. As part of a comparative study, ResNet50 and InceptionV3 were also evaluated, achieving accuracies of 95.83% and 93.75%, respectively.