Interpretable Computer Vision for Brain Tumor Diagnosis: A Convolutional Variational Attention Transform Approach.

Abstract

Brain tumors are among significant health problems in the world, which needs early and proper diagnosis to be treated.  Conventional MRI interpretation by radiologists is a slow and inaccurate process, yet deep learning models, in spite of their high accuracy, lack black-box transparency, which damages the trust and acceptance of clinicians.  Data privacy regulations, such as GDPR and HIPAA, make it difficult to have a centralized training of various datasets, limiting the extrapolation of the models in constrained resources. This paper proposes ConVAT, a novel interpretable deep learning model integrating ResNet50 as the convolutional backbone with variational autoencoders for latent representations and transformer-based attention for spatial-contextual focus. Trained on three diverse MRI datasets—BraTS-Africa (binary classification: glioma vs. other neoplasms, 95 patients), Figshare (multi-class: glioma, meningioma, pituitary tumor, 233 patients), and Brain Tumor Progression (progression stages: early, mid, late, 20 patients with 8798 images)—ConVAT incorporates explainable AI techniques including Smooth Grad-CAM, Score-CAM, and attention visualizations to enhance transparency. Over 50 epochs, the model achieves 98.7%, 99.70% and 99.11 % test accuracy across all datasets, with weighted precision, recall, F1-score, and specificity of 98.7%, 99.70% and 99.11 %, outperforming baselines in generalization and interpretability. Comparative analysis reveals Score-CAM as the most reliable XAI method for clinical validation. By addressing opacity, privacy, and efficiency, ConVAT can facilitate the adoption of AI in healthcare where its use can increase diagnostic accuracy and radiologist trust.

Index Terms—Brain tumor classification, interpretable AI, convolutional variational attention transform, ResNet50, explainable AI, MRI imaging.