Study on the Evaluation of Hip Dysplasia by Measuring the Lateral Center-Edge Angle of Hip Joint on X-Ray Using Deep Learning Algorithm.

Objective: To study the effectiveness and value of using a deep learning algorithm to measure the lateral center-edge angle (LCEA) of the hip joint on X-ray images for the evaluation of hip dysplasia. Methods: This retrospective study included 231 patients (462 hips) undergoing bilateral hip X-rays from February 2023 to February 2024. Two radiologists annotated key acetabular landmarks and femoral contours for model training. A deep learning model performed automatic LCEA measurements, which were compared to manual measurements by radiologists. Consistency was assessed using intraclass correlation coefficient (ICC), and diagnostic performance was evaluated with ROC analysis. Results: A total of 462 hips were measured. There was no statistically significant difference between manual and automated measurements of left and right LCEA (P>0.05). The ICC measurements for manual and automated LCEA for the left and right hips were 0.936 and 0.902, respectively, with r-values of 0.929 and 0.913 (P<0.05). A total of 44 hips were diagnosed with hip dysplasia and 56 with borderline hip dysplasia. The sensitivity of automated measurements for diagnosing hip dysplasia was 88.64% (39/44), with an accuracy of 95.67% (442/462); the sensitivity for diagnosing borderline dysplasia was 66.07% (37/56), with an accuracy of 91.56% (423/462). The area under the curve (AUC) for diagnosing hip dysplasia using automated LCEA measurements (threshold of 25.24°) was 0.917 (P<0.05). Conclusion: The deep learning algorithm for measuring bilateral hip LCEA on X-rays has good efficacy in diagnosing hip dysplasia, with an AUC of up to 0.917.