Structure-enhanced local phase filtering using L₀ gradient minimization for efficient semiautomated knee magnetic...

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Radiology

Background:

• Accurate segmentation of bone surfaces from MRI is crucial for knee osteoarthritis assessment and surgical planning.
• Challenges in MRI segmentation include low contrast, noise, bias fields, and partial volume effects, hindering precise bone delineation.

Purpose of the Study:

• To develop and validate a framework for enhancing knee MRI scans to improve bone surface segmentation.
• To achieve high-contrast bone images for quantitative analysis and patient-specific implant design.

Main Methods:

• A multi-stage framework involving contrast enhancement using relative total variation regularization and sparse gradient counting.
• Incorporation of local phase information for intensity-invariant contrast enhancement.
• Segmentation of enhanced images using a fast random walker algorithm.

Main Results:

• The framework successfully enhances contrast between bone and surrounding tissues in knee MRI.
• Validation on 20 clinical knee MRI volumes demonstrated a mean Dice similarity coefficient of 0.949 compared to expert segmentation.
• The method effectively addresses noise and low contrast issues inherent in MRI data.

Conclusions:

• The proposed framework significantly improves the quality of bone surface segmentation in knee MRI.
• This enhanced segmentation accuracy has direct implications for improved quantitative measurements and surgical planning in knee arthroplasty.
• The developed method offers a robust solution for overcoming common MRI segmentation challenges.