Automated Femoral, Acetabular, and Global Offset Measurements on Anteroposterior Pelvis Radiographs: A Deep-Learning Analysistext_resourcesINTRODUCTION: Measurement of femoral offset is critical in total hip arthroplasty (THA) for guiding native anatomy restoration. However, measurements can be time-consuming and measurer-dependent, creating obstacles for large cohort measurement analyses. We aim to create an objective and reliable offset measurement algorithm using deep learning. METHODS: A total of 500 patients with baseline and 96-month follow-up pelvis radiographs from the Osteoarthritis Initiative (OAI) were included in algorithm training. Radiographs were segmented with identification of the teardrop, femoral head, implant head, and femoral diaphysis below the lesser trochanter. The data were split into training (80%) and validation sets (20%). A U-Net convolutional neural network was trained to identify the landmarks and optimized using the multi-class Dice coefficient metric. Femoral axis and femoral/implant head center of rotation were calculated with the model predictions, and measurements of offset were compared against two trained readers on an independent testing cohort. RESULTS: The optimized model had a Dice coefficient of 0.96 and a foreground mask accuracy of 96.2%. The model measured femoral, acetabular, and global offset on both limbs at a rate of 1.67 s/image. On an independent cohort of 90 hips, the ICC between readers and the deep learning algorithm was 0.86 [95% CI: 0.80–0.91] for femoral offset, 0.87 [95% CI: 0.78–0.91] for acetabular offset, and 0.94 [95% CI: 0.91–0.96] for global offset. When applied to the entire OAI cohort (n = 4188), all images were processed and measured in 116 minutes, and all relevant anatomical features (femoral axis, implant/femoral center of rotation, inter-teardrop line) were correctly calculated in 83.0% of images (n = 3,477). DISCUSSION AND CONCLUSION: We report the development of an accurate and rapid offset measurement model using deep learning that can be applied pre- and post- THA. Future work will involve the external validation of this model with a patient cohort from a different institution using both 2D and 3D imaging.Jack C. Casey, Seong J. Jang, Billy Kim, Kyle Kunze, Chris Anderson, David J. Mayman, Seth A. Jerabek, Jonathan M. Vigdorchik, Peter K. SculcoAuthorBrown University. Alpert Medical School. Scholarly Concentration Program. Medical Technology, Innovation and Entrepreneurshipresearch programOrthopedic surgeryEnglish2023Medical Technology, Innovation and EntrepreneurshipAll rights reservedIn CopyrightAll Rights Reserved