JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

Predicting rapid progression in knee osteoarthritis: a novel and interpretable automated machine learning approach, with specific focus on young patients and early disease

  • 0Department of Surgery, University of Cambridge, Cambridge, UK sc2257@cam.ac.uk.
Annals of the Rheumatic Diseases +

|

September 5, 2024

|

September 5, 2024

View abstract on PubMed

Summary

This summary is machine-generated.

We developed an automated machine learning tool to predict rapid knee osteoarthritis progression. This reliable, transparent tool aids in patient stratification for osteoarthritis (OA) treatment development and clinical trials.

Area Of Science

  • Biomedical Engineering
  • Data Science in Healthcare
  • Rheumatology

Background

  • Osteoarthritis (OA) poses a significant challenge for patient stratification in treatment development and clinical trials.
  • Predicting rapid knee OA progression is crucial for personalized medicine and efficient trial recruitment.

Purpose Of The Study

  • To develop and validate an automated machine learning (autoML) tool for predicting rapid knee OA progression over a two-year period.
  • To facilitate patient stratification for novel OA treatment development and clinical trial recruitment.

Main Methods

  • Developed autoML models integrating clinical, biochemical, X-ray, and MRI data from the OA Initiative datasets.
  • Validated models using two outcome definitions (multiclass and binary) and external datasets.
  • Identified key predictors of progression and conducted subgroup analyses by age, sex, and ethnicity.

Main Results

  • Simpler models using only clinical variables achieved robust external validation (AUC-PRC 0.727 for multiclass, 0.764 for binary).
  • Models performed best in early-stage OA patients and those younger than 60.
  • Patient-reported outcomes and MRI features were key predictors; personalized prediction web applications were developed.

Conclusions

  • The novel autoML tool offers transparency and reliability in predicting rapid knee OA progression.
  • This tool is poised for clinical acceptance, distinguishing itself from 'black-box' methods.
  • Enables effective implementation in clinical practice for improved OA management.

Keywords: