AI-BASED CLASSIFICATION AND DETECTION OF BRAIN TUMORS IN HEALTHCARE IMAGING DATA

Abstract

This study proposes an AI-driven framework for brain tumour detection using explainable AI (XAI) and deep learning to enhance diagnostic accuracy and transparency in healthcare imaging. The methodology integrates cloud-hosted MRI datasets, pre-processed via Median Filtering and Contrast-Limited Adaptive Histogram Equalization (CLAHE), with an EfficientNet-based classifier optimized through compound scaling and auxiliary classifiers. SHAP (Shapley Additive Explanations) provides interpretable insights into model decisions, ensuring clinical trust. Results demonstrate high accuracy (99.0%), though lower precision (90.94%) and recall (90.55%) highlight challenges in class balance. The ROC curve’s marginal AUC (0.5439) underscores limitations in tumour vs. healthy tissue discrimination. Despite computational demands and protocol dependencies, the framework offers a scalable, privacy-preserving solution for clinical deployment, bridging gaps between AI performance and interpretability in oncology.