Radiomic features of infrapatellar fat pad are associated with knee symptoms and radiographic post-traumatic osteoarthritis at 10+ years after anterior cruciate ligament reconstruction

Abstract

Objective

The infrapatellar fat pad (IPFP) has been identified as a potential agent in joint degeneration leading to post-traumatic osteoarthritis (PTOA) in patients suffering from anterior cruciate ligament (ACL) injury. We leveraged machine learning and radiomics methods on knee MRI taken at ten-year follow-up post-ACL reconstruction to associate IPFP with knee symptoms and radiographic PTOA.

Design

In this cross-sectional study, the multi-site NIH-funded MOON nested Onsite cohort was followed up at ten years to obtain 3D MRI radiomics and patient-reported outcome measures (PROM). We identified the features with two radiomics-based classifiers that can detect, respectively, knee symptoms based on PROM data or radiographic PTOA based on Kellgren-Lawrence grade.

Results

We identified 29 radiomics features describing IPFP texture heterogeneity, volume, and signal intensity. For knee symptom detection, models constructed from radiomics achieved an AUROC of 0.76 [95 % CI, 0.65, 0.87], and 0.74 on cross-validation and the test set, respectively. For radiographic PTOA detection, models combining radiomics with clinical features achieved an AUROC of 0.82 [95 % CI, 0.74, 0.92] and 0.79 on cross-validation and the test set, respectively. Increased IPFP texture heterogeneity, larger volume, and increased signal intensity were linked to higher likelihood of knee symptoms and radiographic PTOA.

Conclusion

Radiomics features describing IPFP intensity, morphology, and texture achieve fair to moderate performance in discriminating PTOA-positive from PTOA-negative patients, defined either symptomatically or radiographically. These features describe the relationship between the IPFP and PTOA and are candidates for prognostic models or diagnostic scores that would link knee imaging to patient symptoms.