Brain Tumor Classification and Detection by Explainable AI, Xception and EfficientNet Models for Improving Performance Metrices

The brain tissue develops tumors through abnormal cellular growth that either remains benign or transforms into malignancy. The accurate detection together with proper classification of brain tumors leads to decisions about treatment options and enhances patient survival potential. MRI scan interpretation through radiological assessment for brain tumor diagnosis currently depends on lengthy manual review by experts who face human interpretive bias along with errors. The current diagnostic obstacles emphasize why automated precise and interpretable medical diagnostics systems need development. This research develops an explainable deep learning system by applying transfer learning models Xception and EfficientNetB0 to perform multiclass brain tumor identification through MRI image analysis. The proposed framework handles deep learning system black boxes through Grad-CAM (Gradient-weighted Class Activation Mapping) implementation as an explainable AI (XAI) technique. Part of this framework shows visual explanations about model decisions through MRI regions that most heavily influence classification results. The proposed method succeeds in both achieving accurate diagnoses and delivering diagnostic explanations to healthcare practitioners thus creating conditions for AI-driven diagnosis acceptance. Medical professionals utilize the framework because it grants them explicit information about how predictions are generated to enable better decision-making. This tool possesses the capability to transform brain tumor diagnosis by delivering customized time-sensitive treatments which enhance patient results.