Machine learning to predict incident radiographic knee osteoarthritis over 8 years using combined MR imaging features, demographics, and clinical factors: Data from the Osteoarthritis Initiative
Osteoarthritis and Cartilage Nov 23, 2021
Joseph GB, McCulloch CE, Nevitt MC, et al. - This study provides a 10-predictor model that enables good prediction of incident radiographic osteoarthritis (OA) over 8 years. This model includes MRI parameters together with demographics, symptoms, muscle, and physical activity scores.
From the Osteoarthritis Initiative database, 1044 individuals with baseline Kellgren Lawrence grade 0-1 in the right knee were included.
Three models were created and compared: Model 1: 112 predictors based on OA risk factors; Model 2: top 10 predictors based on feature importance score from Model 1 and clinical relevance; Model 3: Model 2 without the imaging predictors.
The 10-predictor model (Model 2, that incorporates cartilage and meniscus Whole-Organ Magnetic Resonance Imaging Scores and cartilage T 2 ) afforded a AUC of 0.772 which is slightly lower than the model with 112 predictors (Model 1: AUC=0.792).
The 10-predictor model also showed a significantly higher AUC vs the model without MR imaging predictors (Model 3, AUC=0.669, p=0.011).
-
Exclusive Write-ups & Webinars by KOLs
-
Daily Quiz by specialty
-
Paid Market Research Surveys
-
Case discussions, News & Journals' summaries