Prediction of knee pain improvement over two years for knee osteoarthritis using a dynamic nomogram based on MRI-derived radiomics: a proof-of-concept study

Abstract

Objectives: To develop and validate a nomogram to detect improved knee pain in osteoarthritis (OA) by integrating magnetic resonance imaging (MRI) radiomics signature of subchondral bone and clinical characteristics. Methods: Participants were selected from the Vitamin D Effects on Osteoarthritis (VIDEO) study. The primary outcome was 20% improvement of knee pain score over 2 years in participants administrated either vitamin D or placebo. Radiomics features of subchondral bone and clinical characteristics from 216 participants were extracted and analyzed. The participants were randomly split into the training and validation cohorts at a ratio of 8:2. Least absolute shrinkage and selection operator (LASSO) regression was used to select features and generate radiomics signatures. The optimal radiomics signature and clinical indicators were fitted into a nomogram using multivariable logistic regression model. Results: The nomogram showed favorable discrimination performance [AUCtraining, 0.79 (95% CI: 0.72–0.79), AUCvalidation, 0.83 (95% CI: 0.70–0.96)] as well as a good calibration. Additional contributing value of fusion radiomics signature to the nomogram was statistically significant (NRI, 0.23; IDI, 0.14, P < 0.001 in training cohort and NRI, 0.29; IDI, 0.18, P < 0.05 in validating cohort). Decision curve analysis confirmed the clinical usefulness of nomogram. Conclusion: The radiomics-based nomogram comprising the MR radiomics signature and clinical variables achieves a favorable predictive efficacy and accuracy in differentiating improvement in knee pain among OA patients. This proof-of-concept study provides a promising way to predict clinically meaningful outcomes.