Download PDFOpen PDF in browserSolutions for Lower Limb Misalignment: A Segmentation-Guided Coordinate Regression Approach for Landmark Detection and Automatic Measurement6 pages•Published: December 17, 2024AbstractTraditionally, surgeons measure lower limb deformities manually by assessing angles between axes drawn on full lower limb X-rays connecting specific landmarks. This process is considered cumbersome and subject to the surgeon's expertise. Our study aims to alleviate the manual detection of landmarks while enhancing the assessment of lower limb malalignment through an innovative approach that combines coordinate regression and landmark segmentation. While various deep learning solutions exist, our method differs by using landmark segmentation to indicate the possible location of the landmarks; this information is combined with the X-rays to estimate the position of the landmarks via coordinate regression. We named this deep learning architecture segmentation-guided regression.To address the performance of our proposed approach, we evaluated the detection errors for eight landmarks and measured five malalignment metrics. We also compare our approach against landmark regression and landmark segmentation. While landmark segmentation achieved accurate landmark identification, it faced challenges in malalignment measurement due to incorrectly detected landmarks. On the other hand, regression had no failed detections but exhibited lower landmark detection accuracy. Our segmentation-guided regression showed a balance, ensuring no incorrect landmark detections, improved landmark accuracy, and precise malalignment quantification. By encouraging the coordinate regression network to focus on specific areas through segmentation guidance, our method positions landmarks more accurately and effectively measures malalignment. Consequently, our approach provides surgeons with a reliable tool for comprehensive lower limb malalignment assessment, combining the strengths of coordinate regression and landmark segmentation. Keyphrases: accuracy, artificial intelligence, automation, deep learning, lower limb misalignment, x rays In: Joshua W Giles and Aziliz Guezou-Philippe (editors). Proceedings of The 24th Annual Meeting of the International Society for Computer Assisted Orthopaedic Surgery, vol 7, pages 1-6.
|
