The following is a summary of the “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,” published in the February 2023 issue of Osteoarthritis and Cartilage by Lin, et al.
The goal was to create and verify a nomogram that uses clinical characteristics and the magnetic resonance imaging (MRI) radiomics signature of subchondral bone to identify patients with improved knee pain due to osteoarthritis (OA). This study recruited people who were taking part in the Vitamin D Effects on Osteoarthritis (VIDEO) study. Participants were randomised to receive either vitamin D or a placebo, and the primary outcome was a 20% reduction in knee pain score over 2 years. Radiomics subchondral bone features and clinical characteristics were extracted and analyzed from 216 participants.
A ratio of 8:2 was used to divide the participants into training and validation groups randomly. The features were selected and radiomics signatures were generated using least absolute shrinkage and selection operator (LASSO) regression. Using a multivariate logistic regression model, the ideal radiomics signature and clinical indicators were mapped onto a nomogram. There was good calibration and good discrimination performance from the nomogram [AUCtraining, 0.79 (95% CI: 0.72-0.79), AUCvalidation, 0.83 (95% CI: 0.70-0.96)].
The fusion radiomics signature added statistically significant value to the nomogram (NRI, 0.23; IDI, 0.14, P<0.001 in training cohort; NRI, 0.29; IDI, 0.18, P<0.05 in validating cohort). Analysis of decision curves demonstrated nomogram’s clinical value. Improvement in knee pain among OA patients can be reliably predicted using a radiomics-based nomogram that incorporates the MR radiomics signature and clinical variables. This proof-of-concept research suggests a promising approach to predicting clinically meaningful outcomes.