Multi-classifier prediction of knee osteoarthritis progression from incomplete imbalanced longitudinal data

Conventional inclusion criteria used in osteoarthritis clinical trials are not very effective in selecting patients who would benefit the most from a therapy under test. Typically these criteria select majority of patients who show no or limited disease progression during a short evaluation window of the study. As a consequence, less insight on the relative effect of the treatment can be gained from the collected data, and the efforts and resources invested in running the study are not paying off. This could be avoided, if selection criteria were more predictive of the future disease progression. In this article, we formulated the patient selection problem as a multi-class classification task, with classes based on clinically relevant measures of progression (over a time scale typical for clinical trials). Using data from two long-term knee osteoarthritis studies OAI and CHECK, we tested multiple algorithms and learning process configurations (including multi-classifier approaches, cost-sensitive learning, and feature selection), to identify the best performing machine learning models. We examined the behaviour of the best models, with respect to prediction errors and the impact of used features, to confirm their clinical relevance. We found that the model-based selection outperforms the conventional inclusion criteria, reducing by 20-25% the number of patients who show no progression and making the representation of the patient categories more even. This result indicates that our machine learning approach could lead to efficiency improvements in clinical trial design.

Bone Texture Analysis for Prediction of Incident Radio-graphic Hip Osteoarthritis Using Machine Learning: Data from the Cohort Hip and Cohort Knee (CHECK) study

Our aim was to assess the ability of radiography-based bone texture parameters in proximal femur and acetabulum to predict incident radiographic hip osteoarthritis (rHOA) over a 10 years period. Pelvic radiographs from CHECK (Cohort Hip and Cohort Knee) at baseline (987 hips) were analyzed for bone texture using fractal signature analysis in proximal femur and acetabulum. Elastic net (machine learning) was used to predict the incidence of rHOA (Kellgren-Lawrence grade (KL) > 1 or total hip replacement (THR)), joint space narrowing score (JSN, range 0-3), and osteophyte score (OST, range 0-3) after 10 years. Performance of prediction models was assessed using the area under the receiver operating characteristic curve (ROC AUC). Of the 987 hips without rHOA at baseline, 435 (44%) had rHOA at 10-year follow-up. Of the 667 hips with JSN grade 0 at baseline, 471 (71%) had JSN grade > 0 at 10-year follow-up. Of the 613 hips with OST grade 0 at baseline, 526 (86%) had OST grade > 0 at 10-year follow-up. AUCs for the models including age, gender, and body mass index to predict incident rHOA, JSN, and OST were 0.59, 0.54, and 0.51, respectively. The inclusion of bone texture parameters in the models improved the prediction of incident rHOA (ROC AUC 0.66 and 0.71 when baseline KL was also included in the model) and JSN (ROC AUC 0.62), but not incident OST (ROC AUC 0.53). Bone texture analysis provides additional information for predicting incident rHOA or THR over 10 years.