Deep Learning Algorithm of the SPARCC Scoring System in SI Joint MRI.

Area of Science:

• Artificial Intelligence in Medical Imaging
• Radiology and Diagnostic Imaging
• Rheumatology and Musculoskeletal Diseases

Background:

• The Spondyloarthritis Research Consortium of Canada (SPARCC) scoring system is a standard for grading sacroiliitis.
• Accurate sacroiliitis grading is crucial for diagnosing and managing spondyloarthritis.

Purpose of the Study:

• To develop and evaluate a deep learning (DL) pipeline for automated sacroiliitis grading.
• To assess the DL pipeline's performance against the established SPARCC scoring system.

Main Methods:

• A prospective study involving 389 participants with sacroiliitis.
• Development of DL models for segmenting bone marrow edema (BME) and sacroiliac joints on 3-T STIR MRI sequences.
• Comparison of DL pipeline scores with expert human readers using intraclass correlation coefficient (ICC) and Pearson coefficient.

Main Results:

• The DL pipeline achieved high consistency with human readers, with ICC of 0.83 and Pearson coefficient of 0.86.
• High sensitivity (0.83) for BME detection and accuracy (0.90) for SI joint identification were observed.
• Dice coefficients for sacrum and ilium segmentation were 0.82 and 0.80, respectively.

Conclusions:

• The developed deep learning pipeline demonstrates strong performance in grading sacroiliitis based on the SPARCC system.
• This DL approach offers a reliable and consistent method for scoring STIR MRI images in spondyloarthritis patients.
• The findings suggest potential for AI to enhance the efficiency and accuracy of sacroiliitis assessment.