Automatic framework for evaluating osteoarthritic cartilage severity: high-resolution cartilage thickness mapping and...

Area of Science:

• Biomedical imaging
• Radiology
• Osteoarthritis research

Background:

• Current osteoarthritis assessment methods lack scalability and struggle with variable cartilage degeneration.
• Enhanced imaging biomarkers are crucial for osteoarthritis research and clinical trials.

Purpose of the Study:

• To develop and validate an automatic, scalable framework for assessing femoro-tibial osteoarthritic cartilage severity.
• Utilize high-resolution cartilage thickness maps (CTh-Maps) and a novel CTh-Score.

Main Methods:

• A 3D-UNet segmented femoro-tibial bones and cartilages from 4796 subjects in the Osteoarthritis Initiative (OAI) cohort.
• A ResNet model generated CTh-Maps and assigned a CTh-Score (0-100) for cartilage severity.
• Reproducibility, validity against expert grading (MOAKS), and association with OA severity (KL grade) were assessed.

Main Results:

• The framework generated the "OAI CTh-Maps" dataset with excellent reproducibility (ICC > 0.98) for both CTh-Maps and CTh-Score.
• The CTh-Score strongly correlated with expert cartilage loss assessments (r=0.81) and OA severity.
• The CTh-Score demonstrated high sensitivity to change, detecting progression up to 6 years before KL grade changes.

Conclusions:

• CTh-Maps and CTh-Score offer a novel, scalable, reproducible, and sensitive approach to analyzing cartilage in imaging.
• This framework enables earlier OA detection, improved disease monitoring, and identification of therapeutic windows.
• The CTh-Score serves as a valuable biomarker for structural progression in clinical trials.