Segmentation of the human spinal cord

Area of Science:

• Medical Imaging
• Neuroscience
• Radiology

Background:

• Spinal cord segmentation is essential for quantifying atrophy and analyzing white/gray matter (WM/GM) in diseases.
• Spinal cord segmentation lags behind brain segmentation but has advanced significantly with improved MRI sequences.
• Existing techniques vary in complexity and accuracy, necessitating standardized validation.

Purpose of the Study:

• To review existing spinal cord and WM/GM segmentation methods.
• To provide recommendations for validating segmentation techniques.
• To illustrate applications in healthy and pathological spinal cords.

Main Methods:

• Review of intensity-based, surface-based, and image-based segmentation methods.
• Discussion of validation strategies for spinal cord segmentation techniques.
• Illustrative examples of segmentation in clinical applications.

Main Results:

• No single method currently achieves adequate segmentation in all clinical scenarios.
• Robust and automatic segmentation is vital for unbiased, reproducible longitudinal and multicenter studies.
• A combination of approaches and probabilistic shape models are recommended for future development.

Conclusions:

• Advancements in spinal cord MRI have spurred progress in segmentation techniques.
• Standardized frameworks, multiplatform availability, and data sharing are crucial for community benefit.
• Future research should focus on hybrid methods and probabilistic models for comprehensive segmentation.