Graph cut-based segmentation for intervertebral disc in human MRI.

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Computer Vision

Background:

• Accurate segmentation of lumbar intervertebral discs (IVDs) is crucial for diagnosing spinal conditions.
• Existing segmentation methods often require manual input, limiting efficiency and reproducibility.
• Degeneration of IVDs is a common cause of low back pain, necessitating reliable imaging analysis.

Purpose of the Study:

• To develop and validate an automated algorithm for 2D segmentation of lumbar intervertebral discs (IVDs) from T2-weighted sagittal spine Magnetic Resonance Images (MRIs).
• To accurately differentiate between healthy and degenerated IVDs using advanced image processing techniques.
• To provide an efficient and user-independent solution for IVD segmentation.

Main Methods:

• An automated algorithm utilizing the s-t max-flow/min-cut approach on a directed graph constructed from image pixels.
• Preprocessing steps include noise reduction and intensity inhomogeneity correction.
• Automatic seed point initialization using a growing bounding box, eliminating manual selection.
• Incorporation of anatomical knowledge of soft tissues for improved segmentation accuracy.

Main Results:

• The algorithm successfully segmented both healthy and degenerated lumbar intervertebral discs (IVDs).
• Achieved a high segmentation accuracy with a Dice Similarity Coefficient (DSC) of 89% on a dataset of 15 patients.
• Demonstrated efficient performance due to polynomial time complexity, enabling globally optimal solutions.

Conclusions:

• The developed automated algorithm provides an efficient and accurate method for lumbar intervertebral disc (IVD) segmentation from MRI.
• The s-t max-flow/min-cut approach effectively distinguishes IVDs from background, handling both healthy and degenerated cases.
• This user-independent technique holds significant potential for clinical applications in diagnosing spinal pathologies.