Deep Learning-based Multi-Class Brain Abnormality Diagnosis on Bangladeshi MRI Images

Abstract

Brain disorders can impair essential functions like thinking, speech, and movement. Early diagnosis is critical to ensure timely and effective treatment. Magnetic resonance imaging (MRI) is widely used to detect such conditions. Still, manual analysis of MRI scans is often time-intensive and can overlook subtle changes, particularly in the early stages of disease. Identifying the most relevant features and selecting suitable classifiers for optimal performance adds complexity. Recently, deep learning models have become increasingly popular for analyzing medical images. This paper presents an automated diagnostic system using a deep convolutional neural network (CNN) to classify various brain abnormalities. We applied several pre-trained models, including CNN, U-Net, VGG16, ResNet50, MobileNetV3, NasNet, and LeNet, to differentiate MRI scans into Stroke, Brain tumor, and normal categories. The dataset, featuring MRI images from Bangladesh, focuses on prevalent brain disorders in that region. Our approach demonstrated strong classification accuracy in CNN with 97% and can serve as a tool for clinicians to verify their manual interpretations of MRI results.

Keywords

Citations by Year

Entered manually, not live-tracked.