Brain Tumor Classification and Growth-rate Analysis using Deep Learning

Abstract

The recent development in the field of artificial intelligence and digital medical imaging has also disrupted the process of analysis and diagnosis of neurological disorders. The automated examination of brain MRI scans has gained more significance in helping radiologists to diagnose tumors with high accuracy as well as decrease the time and manual work involved in diagnosis. The problem of brain tumor detection and classification is not resolved yet, as each patient has different tumors, and some have different shapes, sizes, and intensity. The proposed project will resolve these problems by implementing deep learning based object detection models to tackle medical image analysis with an automated classifier of brain tumors and their rate of growth with the help of the YOLOv8 deep learning model. Preprocessing of MRI images involves a preprocessing pipeline comprising of resizing and normalization of the aspect ratio to enhance localization of tumors to be used as a bounding box in detecting them. Brain tumors are recognized as four types; glioma, meningioma, pituitary tumor and no tumor (YOLOv8 model). The system provides the results of prediction as the type of tumor, the segmented part of the tumor, and the confidence score. The size of the tumors is estimated by the area identified and the rate of growth is calculated based on the analysis carried out by medical rules. The experimental findings prove that the suggested deep learning model not only offers effective tumor detection and classification but also effective growth evaluation, which makes it appropriate as a clinical decision support system in the field of early diagnosis and regular follow-ups.