Neuroimaging and machine learning fusion for improved brain tumor diagnosis and prognosis

Brain diseases, including tumors, along with mental and neurological disorders, pose a substantial threat to millions of individuals worldwide. Doctors use both structural and functional neuroimaging methods for diagnosing these diseases. Specialist doctors often combine Magnetic Resonance Imaging (MRI) with other neuroimaging techniques to enhance the detection and identification of brain diseases in clinical practice. This is crucial because these modalities give different information that supplement each other, hence enhancing accurate diagnosis. This study proposes a fusion approach integrating MRI with three Machine Learning (ML) classifiers, Convolutional Neural Network (CNN), Random Forest (RF), and Support Vector Machine (SVM), to improve brain tumor classification. Using a publicly available dataset containing 7023 MRI images across four categories (glioma, meningioma, pituitary, and no tumor), the models were trained and evaluated using a standard train-test split. The CNN model achieved the highest accuracy at 99.29%, followed by RF at 99.06% and SVM at 98.36%. While the performance metrics are promising, the study is limited to a single dataset and lacks external validation, which constrains the generalizability of results. Future work should explore clinical integration, federated learning frameworks, and multi-center datasets to enhance real-world applicability.