Radiographic Findings Associated With Mild Hip Dysplasia in 3869 Patients Using a Deep Learning Measurement Tool

Area of Science:

• Orthopedics and Radiology
• Artificial Intelligence in Medicine
• Biomedical Imaging Analysis

Background:

• Hip dysplasia is a primary cause of hip degeneration and the need for total hip arthroplasty (THA).
• Accurate measurement of radiographic angles is crucial for diagnosing hip dysplasia.
• Current methods can be time-consuming and subject to inter-observer variability.

Purpose of the Study:

• To validate a deep learning (DL) algorithm for measuring key hip dysplasia angles.
• To apply the validated DL algorithm to a large patient cohort to determine dysplasia prevalence.
• To assess the relationship between radiographic indices and the risk of THA.

Main Methods:

• A DL algorithm was developed to automatically measure modified lateral center-edge angle (LCEA), Tönnis angle, and modified Sharp angle.
• The algorithm was validated against manual measurements in a cohort of 3869 patients from the Osteoarthritis Initiative.
• Prevalence was analyzed using percentile distributions and established radiographic cutoffs.

Main Results:

• The DL algorithm demonstrated excellent agreement with manual measurements (kappa = 0.78-0.88) and no significant difference in angle measurements.
• Automated measurements of 23,214 angles were completed in 140 minutes.
• Dysplastic hip prevalence varied from 2.5% to 20% based on specific angle cutoffs, with 12.4% of patients showing indices linked to higher THA risk.

Conclusions:

• A validated DL algorithm can efficiently measure hip dysplasia angles.
• The prevalence of hip dysplasia in the studied cohort is significantly influenced by the chosen measurement and threshold.
• The DL algorithm provides a reliable tool for identifying individuals at increased risk for THA due to hip dysplasia.