CerebralNet meets Explainable AI: Brain tumor detection and classification with probabilistic augmentation and a deep learning approach

Brain tumors pose a significant and often life-threatening challenge to human health. Accurate and timely brain tumor detection is crucial for early intervention, improved treatment outcomes, and ultimately, enhanced patient survival and quality of life. Automation is crucial for overcoming the limitations of manual analysis, enabling faster, more objective, and potentially more accurate diagnoses of brain tumors. This study introduces the CerebralNet architecture, a new approach for brain tumor detection and classification that uses a pre-trained MobileNetV2 backbone. Key innovations include the integration of Atrous Spatial Pyramid Pooling and Atrous Convolution blocks, enhancing feature extraction, and capturing multi-scale contextual information which is crucial for accurate brain tumor detection and classification because brain tumors exhibit significant variability in size, shape, and appearance. Furthermore, a new probabilistic image augmentation selection strategy is employed, incorporating 10 augmentation techniques such as Gaussian noise, Gaussian blur, random rotations, and intensity and color variations to simulate real-world imaging artifacts. This approach, combined with the augmented Brain MRI Dataset, significantly improves model robustness and generalizability. Rigorous evaluation on both the imbalanced original (BM) and augmented (ABM) datasets demonstrates exceptional performance, exceeding 91 % accuracy on the BM dataset and achieving 96 % on the ABM dataset. This study also incorporates LIME (Local Interpretable Model-agnostic Explanations) for model interpretability, providing valuable insights into the model’s decision-making process. These findings demonstrate the potential of the new augmentation strategy and Atrous MobileNetV2 architecture to significantly improve the accuracy and reliability of brain tumor detection and classification, paving the way for improved clinical outcomes.