Predictive Application for Pre-Screening and Analysis of Brain Tumor using MRI

Detecting brain tumors with MRI imaging remains an important but resource-intensive work for radiologists, often hampered by time and the inherent subjectivity of manual interpretation. This research proposed an enhanced predictive application that uses machine learning and deep learning technologies to simplify pre-screening and improve clinical procedures for brain tumor analysis. For this, firstly we explored two mostly used machine learning models, Support Vector Machine (SVM) and Naive Bayes, and two deep learning models, VGG-16 and ResNet-50, considering the dataset containing four classes, including Glioma, Meningioma, Pituitary, and no tumor. Among them, ResNet-50 achieved the highest classification accuracy (92.02%), leading it to be included in the prediction application. This improved model allowed doctors to conduct rapid and precise early assessments by enabling both lab assistants and patients to upload images and manage them in the system. Meanwhile, the system was evaluated through a cognitive walk-through approach, incorporating physician expertise into the application’s validation process. The application offers excellent diagnostic reliability and clinical utility by providing an easy-to-use interface for image analysis. As such, this study highlights the potential for AI-enhanced solutions to assess medical expertise, expedite diagnostic workflows, and improve patient management through accurate and efficient pre-screening.