Research Information System
Acta Orthopaedica et Traumatologica Turcica, vol.60, no.1, 2026 (SCI-Expanded, Scopus, TRDizin)
Objective: The anteroposterior (AP) pelvic X-ray is commonly used for assessing conditions affecting the bony pelvis. The objective of this study was to develop new pelvic X-ray assessment software (CalculOrther) to assess AP pelvic X-rays and evaluate its reliability. Methods: CalculOrther was developed in 4 stages. Initially, a dataset comprising pelvic X-rays was generated. During the second stage, the convolutional neural network model was trained to identify anatomical landmarks in the pelvic X-ray images. The Hough transform was used to locate the circle and center of the femoral head in the third stage. The border pixels were generated using mathematical morphological processes, and the requisite angles were measured in the fourth stage. Then manual measurements and the software developed were analyzed with Pearson’s correlation and intraobserver and interobserver correlation coefficients. Subsequently, the mean error and the root mean square error (RMSE) were acquired. Results: The Pearson’s correlation coefficients varied from 0.84 to 0.99 (P <.001). The interobserver and intraobserver correlation coefficients ranged from 0.77 to 0.99 and from 0.75 to 0.94, respectively. The RMSE ranged from 0.31 to 4.38 and the mean error from 0.05 to 2.86. The mean duration for manual measurements was 230 (177-284) seconds and 215 (160-255) seconds, respectively. The software required an average time of 3.18 (2.95-3.52) seconds to make the same measurements. Conclusion: Regarding femoroacetabular impingement and hip dysplasia, artificial intelligence can analyze pelvic radiographs and generate equally accurate results within a shorter duration compared to traditional measuring methods. Level of Evidence: Level 2, Diagnostic Study.