Multiple LREK active contours for knee meniscus ultrasound image segmentation

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Computer Vision

Background:

• Accurate quantification of knee meniscus degeneration and displacement from ultrasound images is crucial.
• This requires simultaneous segmentation of the femoral condyle, meniscus, and tibial plateau.
• Existing segmentation methods may struggle with homogeneous regions and multiple object delineation.

Purpose of the Study:

• To present a novel active contour model for robust image segmentation.
• To introduce a multiple active contour framework for segmenting multiple objects without overlap.
• To evaluate the performance of the proposed framework against existing methods for knee meniscus ultrasound images.

Main Methods:

• Development of a scalable local regional information on expandable kernel (LREK) active contour model.
• Implementation of a strategy to adapt local window size, preventing confinement in homogeneous regions.
• Introduction of a multiple LREK (MLREK) framework for multi-object segmentation, ensuring contour separation.
• Investigation of parameter choices within the MLREK framework.

Main Results:

• The proposed MLREK model demonstrates improved segmentation performance compared to other active contour models.
• Quantitative analysis using Dice coefficient and Hausdorff distance on real knee meniscus ultrasound images.
• Successful segmentation of multiple structures without merging or overlapping boundaries.

Conclusions:

• The MLREK framework offers a potential advancement for the assessment of knee meniscus degeneration and displacement.
• The model's ability to handle multiple object segmentation accurately is highlighted.
• The findings suggest MLREK's utility in clinical applications for knee joint analysis.