(ICWS 2024) PLNCaE: Advanced Framework for Preprocessed Lightweight Neural Networks in Brain Tumor Classification and Explanation

The brain is one of the most complicated parts of the human body. In addition, the human brain was created by a million nervous cells, which are responsible for controlling emotion, action, and every process that manages our body. However, the impact of radiation rays from natural to artificial has a huge effect on human health. In particular, it is a cause of hundreds of fatal diseases in humans. One of these is a brain tumor, which is an overgrowth of abnormal cells in the brain or near it. To resolve this problem, several studies in medical and computer science were published for early detection and treatment of brain tumors. In this research, the Magnetic Resonance Imaging (MRI) dataset was combined from three sources, including Figshare, SARTAJ, and Br35H. Besides, the PLNCaE (Prepro-LiteNeuro Classification and Explanation) framework was proposed with preprocessing progress by removing the extra margins and combining layers to generate a lightweight Convolutional Neural Network (CNN) model. Moreover, SHapley Additive exPlanations (SHAP) were employed to explain the model's predictions in the outcome. Through various experiments, the suggested model reached an impressive test accuracy score of 97.58% in classifying four classes, including normal, pituitary, glioma, and meningioma. In binary classification, the result achieved the highest accuracy of 99.02% between normal and abnormal images.