Spinal cord grey matter segmentation challenge.

Area of Science:

• Medical Imaging
• Neuroscience
• Computer Vision

Background:

• Accurate segmentation of spinal cord grey and white matter is essential for tissue-specific analysis in magnetic resonance imaging (MRI).
• While automated methods excel at cervical cord cross-sectional area measurement, grey matter segmentation remains challenging due to its small size and complex shape.
• Ongoing research worldwide focuses on improving grey matter segmentation techniques for spinal cord MRI.

Purpose of the Study:

• To organize a grey matter spinal cord segmentation challenge to evaluate various automated methods.
• To characterize the state-of-the-art in spinal cord grey matter segmentation using a standardized multi-center, multi-vendor dataset.
• To identify opportunities for future advancements in spinal cord grey matter segmentation algorithms.

Main Methods:

• A challenge was organized involving six independent spinal cord grey matter segmentation methods.
• A multi-center, multi-vendor dataset acquired with distinct 3D gradient-echo sequences was used for evaluation.
• Performance of automated methods was compared against manual segmentation, considered the current gold-standard.

Main Results:

• All participating algorithms demonstrated good overall performance in identifying the characteristic grey matter 'butterfly' shape.
• Variable performance was observed across different segmentation quality metrics among the tested methods.
• The dataset and challenge website remain publicly available for ongoing research and new submissions.

Conclusions:

• The challenge successfully characterized the current state-of-the-art in spinal cord grey matter segmentation.
• While current methods show promise, variability in performance metrics indicates room for significant improvement.
• Further research and development are needed to enhance the accuracy and reliability of automated spinal cord grey matter segmentation.