Multi Pathway 3D CNN Wit Conditional Random Field for Automated Segmentation of Multiple Sclerosis Lesions in MRI

This project presents a robust automated framework for the segmentation of Multiple Sclerosis (MS) lesions in brain MRI scans, combining the strengths of deep learning and probabilistic graphical models. We introduce a multi-pathway 3D Convolutional Neural Network (3D CNN) architecture, which captures both local and global contextual information through multiple receptive fields. To further enhance boundary refinement and reduce false positives, a Conditional Random Field (CRF) is applied as a post-processing step. Additionally, comparative evaluations are performed using U-Net++ and Swin Transformer-based segmentation models to benchmark performance across different deep learning paradigms. The web-based system is developed using Flask, integrated with a responsive frontend using HTML and CSS for user-friendly lesion visualization and interaction. Our pipeline demonstrates improved lesion detection accuracy and spatial consistency, making it suitable for clinical decision support in neuroimaging diagnostics.

Keywords:
3D CNN, U-Net++, Swin Transformer, Multiple Sclerosis, MRI Segmentation, Conditional Random Field, Flask, Deep Learning, Neuroimaging, Automated Lesion Detection.