A Deep Learning Tool for Minimum Joint Space Width Calculation on Antero-posterior Knee Radiographs.

Area of Science:

• Radiology
• Artificial Intelligence
• Medical Imaging

Background:

• Minimum joint space width (mJSW) is a key quantitative metric for assessing knee osteoarthritis (OA) progression.
• Accurate mJSW measurement is crucial for both native knees and those that have undergone arthroplasty.

Purpose of the Study:

• To develop an automated algorithm for measuring mJSW in the medial and lateral compartments of the knee.
• The algorithm is designed to be flexible for use on both native and post-arthroplasty knee images.

Main Methods:

• An end-to-end algorithm combining deep learning segmentation and computer vision was developed.
• A deep learning model was trained and validated using 583 segmented images.
• A computer vision algorithm was developed and validated using 330 independent ground truth measurements.

Main Results:

• The segmentation model achieved an average Dice score of 0.92.
• The algorithm demonstrated a mean absolute error of 0.85 ± 1.20 mm compared to human measurements.
• The algorithm showed minimal bias, with a mean error of 0.019 mm, and 73.2% of measurements within 1 mm of human assessments.

Conclusions:

• An automated, AI-based algorithm for measuring knee mJSW from anteroposterior radiographs has been successfully developed and validated.
• This tool can significantly streamline population-level clinical research on knee joint natural history.
• It enables clinicians to efficiently quantify radiographic arthritis progression for individual patients over time.