CartiSurface: implicit surface reconstruction for anatomically-aware cartilage thickness mapping in knee MRI

Area of Science:

• Biomedical imaging
• Medical image analysis
• Computational anatomy

Background:

• Quantitative analysis of knee cartilage thickness is crucial for early osteoarthritis (OA) diagnosis and monitoring.
• Current segmentation methods struggle with noisy, inconsistent thickness maps from clinical MRI scans.

Purpose of the Study:

• To introduce CartiSurface, a novel framework for robust knee cartilage thickness estimation.
• To improve anatomical faithfulness and accuracy in cartilage thickness mapping, especially for low-resolution MRI.

Main Methods:

• Developed an implicit surface reconstruction framework learning a signed distance function (SDF) between bone surfaces.
• Incorporated geometric priors (spacing, parallelism, smoothness) via a dedicated loss function.
• Method does not require explicit voxel-wise cartilage labels, enhancing robustness.

Main Results:

• CartiSurface outperformed state-of-the-art methods on the OAI dataset in accuracy and surface regularity.
• Demonstrated robustness to input variability and resolution degradation.
• Successfully captured focal cartilage thinning and maintained surface continuity.

Conclusions:

• CartiSurface offers a clinically viable tool for early OA detection and longitudinal monitoring.
• The framework provides anatomically faithful and continuous cartilage thickness maps.
• Enables advanced biomechanical modeling for knee joint health assessment.