Diagnosing Brain Tumors: An Artificial Intelligence Modeling Approach

Abstract

Brain tumour diagnosis poses significant challenges due to the complexity and heterogeneity of tumour characteristics. This study introduces an innovative deep-learning approach that leverages the complementary strengths of weighted magnetic resonance imaging and advanced deep-learning techniques to enhance the accuracy and reliability of brain tumour diagnosis. Historically, diverse methodologies have been employed, often centring on categorizing imaging modalities into binary distinctions, either cancerous versus non-cancerous images or discerning between benign and malignant tumours. In contrast, this study aims to classify multi-class malignant tumours into their specific categories with optimal precision. The proposed research centres on the differentiation of meningioma, glioma, and pituitary adenoma brain tumours, representing the three primary types of malignant brain tumours, along with non-tumorous MRI scans. This study employs a Convolutional Neural Network (CNN) and a Variational Autoencoder for the detection of brain tumours in Magnetic Resonance Imaging (MRI) images. Hyperparameter tuning for the CNN is achieved through a Grid search. The highest-performing model exhibits an accuracy of 95.25%, the other model reaches a peak accuracy of almost 100% in diagnosing cases without tumours.