The Feasibility and Performance of Total Hip Replacement Prediction Deep Learning Algorithm with Real World Data.

Area of Science:

• Orthopedics
• Artificial Intelligence
• Medical Imaging

Background:

• Hip degenerative disorder is a leading cause of total hip replacement (THR) in geriatric patients.
• Optimal surgical timing for THR is critical for successful post-operative recovery.
• Deep learning (DL) shows promise in medical image analysis for predicting conditions like THR.

Purpose of the Study:

• To develop and validate a sequential two-stage deep learning algorithm for predicting the need for THR within three months.
• To assess the algorithm's performance using real-world data (RWD) from plain pelvic radiography (PXR).

Main Methods:

• A sequential two-stage deep learning algorithm was designed to predict THR necessity from PXR.
• Real-world data comprising 3766 PXRs from 2018-2019 was collected for validation.

Main Results:

• The algorithm achieved high accuracy (0.9633), sensitivity (0.9450), specificity (1.000), and precision (1.000).
• Key performance metrics included a negative predictive value of 0.9009 and an F1 score of 0.9717.
• The area under the curve was 0.972, indicating strong predictive capability.

Conclusions:

• The developed DL algorithm offers an accurate and reliable method for detecting hip degeneration and predicting THR.
• RWD validation confirms the algorithm's utility, potentially saving time and healthcare costs.