Convolutional Neural Network and Channel Attention Mechanism for Multiclass Brain Tumor Classification

The complexity of brain tumors highlights the critical need for advanced computer-aided diagnosis (CAD) tools to support surgeons in clinical decision-making and improve patient outcomes. This paper introduces a novel deep learning model for the multiclass classification of brain tumors using magnetic resonance imaging (MRI), offering significant advancements in feature extraction and classification accuracy. The proposed model comprises three key components: (1) a fine-tuned EfficientNetB7 convolutional neural network (CNN), adapted through transfer learning by freezing the initial layers and retraining subsequent layers to optimize feature extraction from MR images; (2) a channel attention module that refines extracted feature maps, emphasizing essential features for accurate tumor detection; and (3) a fully connected classifier, optimized through grid search, to achieve precise multiclass tumor classification. Additionally, hyperparameter tuning and data augmentation techniques enhance generalization and model robustness. Experimental results confirm the model’s superior performance, outperforming recent approaches in multiclass and binary classification scenarios, underscoring its potential to advance brain tumor diagnosis and treatment.