AI automated grid placement in the OMERACT knee inflammation MRI scoring system (KIMRISS) for bone marrow lesion...

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Medicine
• Rheumatology and Osteoarthritis Research

Background:

• Accurate quantification of bone marrow lesions (BMLs) is crucial for assessing osteoarthritis progression.
• The Knee Inflammation MRI Scoring System (KIMRISS) is a standardized method for BML scoring.
• Manual grid placement in KIMRISS can be time-consuming and subject to inter-reader variability.

Purpose of the Study:

• To validate the reliability and feasibility of an AI-automated grid placement method for BML scoring within the KIMRISS framework.
• To assess the performance of automated grid placement against expert manual placement using the OMERACT Filter criteria.
• To evaluate the usability and efficiency of the AI-automated system.

Main Methods:

• Eleven musculoskeletal radiology experts evaluated 40 MRI datasets.
• Manual and AI-automated grid placements were compared using spatial similarity metrics (Dice Coefficient) and agreement metrics (intraclass correlation coefficients).
• Feasibility was assessed using the System Usability Scale (SUS).

Main Results:

• Automated and manual grids showed strong spatial similarity (mean femur Dice Coefficient = 0.78).
• High agreement was observed between KIMRISS BML scores derived from automated and manual grids (mean ICCs of 0.86 and 0.89 for baseline and change scores).
• The AI system achieved moderate usability scores (mean SUS = 66.1).

Conclusions:

• AI-automated grid placement is a reliable and feasible enhancement to the KIMRISS.
• This automated approach has the potential to improve the ease and reproducibility of BML quantification in osteoarthritis research and clinical trials.
• The findings support the integration of AI tools to streamline MRI-based osteoarthritis assessment.