Interpretable and Efficient Brain Tumor Detection using Efficient Net Enhanced with CBAM and Grad-CAM Analysis

Abstract

Accurate and timely identification of brain tumors from Magnetic Resonance Imaging (MRI) plays a crucial role in clinical decision-making and patient management. Although deep convolutional neural networks (CNNs) have demonstrated remarkable success in medical image classification, their increasing architectural complexity often leads to high computational demands and limited interpretability, which can hinder their adoption in real-world clinical settings. In this study, we present an efficient and interpretable deep learning framework for binary brain tumor classification (tumor versus non-tumor) that combines an EfficientNet backbone with the Convolutional Block Attention Module (CBAM), complemented by Gradient-weighted Class Activation Mapping (Grad-CAM) for model explainability. EfficientNet is employed due to its parameter-efficient design, which relies on compound scaling to achieve a balanced expansion of network depth, width, and input resolution, resulting in substantially lighter models compared to traditional architectures such as VGG or ResNet. The incorporation of CBAM further refines feature learning by sequentially applying channel-wise and spatial attention mechanisms, enabling the network to emphasize diagnostically relevant regions while minimizing the influence of non-informative background structures. Experimental evaluation on the Brain Tumor MRI dataset demonstrates that the proposed EfficientNet–CBAM architecture attains high classification performance, achieving accuracy and F1-score values of up to 99.5%, while maintaining a favorable balance between computational efficiency and diagnostic accuracy relative to existing approaches. Furthermore, the use of Grad-CAM produces intuitive heatmaps that highlight tumor-associated regions influencing the model’s predictions, thereby enhancing transparency, supporting error analysis, and fostering clinical trust. Overall, the proposed framework offers a robust and explainable deep learning solution that effectively aligns high-performance image analysis with the practical requirements of automated medical diagnosis.