MSEF-Net: Multi-scale edge fusion network for lumbosacral plexus segmentation with MR image

Area of Science:

• Medical Imaging
• Neurosurgery
• Artificial Intelligence

Background:

• Spinal nerve damage causes significant disability and paralysis.
• Accurate segmentation of the lumbosacral plexus in MRI is crucial for diagnosing spinal nerve lesions and guiding surgery.
• The complex structure and low contrast of the lumbosacral plexus present challenges for precise edge delineation.

Purpose of the Study:

• To develop an advanced deep learning model for improved lumbosacral plexus segmentation in MRI.
• To enhance the accuracy of edge feature extraction and multi-scale feature fusion for complex anatomical structures.
• To provide a more effective tool for computer-aided diagnosis and surgical planning related to spinal nerve lesions.

Main Methods:

• Proposed a novel Multi-Scale Edge Fusion Network (MSEF-Net) incorporating an edge feature fusion module (EFFM) and an adaptive multi-scale fusion module (AMSF).
• The EFFM combines Sobel operator edge detection with an edge-guided attention module (EAM) to highlight edge structures.
• The AMSF adaptively fuses multi-scale feature maps within the network's decoder.

Main Results:

• The MSEF-Net demonstrated superior performance in lumbosacral plexus segmentation on a dataset of 2848 spinal MRI images from 89 patients.
• Experimental results confirmed the effectiveness of the proposed network compared to existing state-of-the-art segmentation methods.
• The method successfully addressed challenges related to complex structures and low contrast in MRI scans.

Conclusions:

• The MSEF-Net is an effective deep learning approach for accurate lumbosacral plexus segmentation using MRI.
• This improved segmentation capability can enhance computer-aided diagnosis and surgical interventions for spinal nerve conditions.
• The proposed network architecture offers a promising solution for segmenting challenging anatomical regions in medical imaging.