An Effective Deep CNN Model for Multiclass Brain Tumor Detection Using MRI Images and SHAP Explainability

Abstract

In today's health system, clinical diagnosis has taken on a crucial role. Brain cancer, which is the most serious illness and the main cause of death worldwide, is a significant area of study in the field of medical imaging. Brain cancers must be categorized and found to evaluate the tumors and choose the most appropriate course of treatment for each class. To find brain tumors, numerous imaging modalities are used. However, MRI is frequently utilized since it produces superior images and uses no ionizing radiation. Machine learning's area of deep learning (DL) lately displayed impressive performance, particularly in classification and segmentation issues. We offer an explanation-driven Deep Learning model for the prediction of discrete subtypes of brain tumors (meningioma, glioma, and pituitary) utilizing a brain tumor MRI imaging dataset employing an EfficientNetB0 convolutional neural network (CNN) and Shapley additive explanation (SHAP). The proposed brain tumor detection and classification method surpasses previous methods both visually and numerically, according to an experimental investigation, and achieves an accuracy of 99.84%. The eXplainable Artificial Intelligence (XAI) approach (SHAP) is then used to explain the outcome.