No abstract

No abstract

better than the base model and comparable to KL grade (Table 1, Figure1). Finally, all the results consistently improved when a regularized model was applied. Applying regularized logistic regression developed in a machine learning fashion, showed excellent predictive performance of the US features for all models in which these were included (Table 1). Conclusions: Baseline US features could predict future TKR cases comparably well as conventional plain radiography, scored with KL grading, and better than the base model which included clinical covariates only. Finally, the predictive performance of all the built models was improved when a regularized model was used for prediction.

No abstract