Preoperative Assessment of Vessel-to-acetabular Rim Distances in Non-contrast CT Images for Total Hip Arthroplasty

Yingdong CHEN, Mazen SOUFI, Keisuke UEMURA, Yoshito OTAKE, Masaki TAKAO, Shinichi IWAKOSHI, Toshihiro TANAKA, Nobuhiko SUGANO, Yoshinobu SATO
Vol. 13 (2024) p. 176-188

Purpose: Blood vessel injuries during total hip arthroplasty (THA) pose life-threatening risks. This study aimed at creating a preoperative approach for evaluating vessel-to-acetabular rim distances for surgical planning of THA.

Methods: From non-contrast CT images, the pelvis and blood vessels including the external iliac artery and vein were automatically segmented. Surface points on the vessels and the acetabular rim were used to assess distance within high-risk locations. Thirty-six non-contrast CT images of patients with hip osteoarthritis (OA) and 18 pairs of contrast-enhanced/non-contrast CT images of patients with non-hip OA were used for internal training/validation, and 10 non-hip OA CT images collected from a public database were used for external validation. The ground truth (GT) labels were manually annotated for the internal hip OA database, whereas the arterial GT labels for the internal non-hip OA database were constructed using registered contrast-enhanced CT, in which the arteries were clearly imaged. The Dice coefficient (DC) and average symmetric surface distance (ASD) were used to evaluate segmentation accuracy. The accuracy of distance assessment was assessed using mean absolute error (MAE) and Pearson correlation coefficient (PCC).

Results: Using a 3D nnU-Net model, DCs for artery and vein were 0.877 and 0.892, respectively, in internal hip OA patients, 0.901 and 0.909 in internal non-hip OA patients, and 0.846 and 0.858 in external non-hip OA patients. PCCs between the GT and auto-segmentation-based distances were larger than 0.97, with a mean MAE smaller than 0.5 mm in the high-risk location.

Conclusions: The study proposes a preoperative approach for vessel-to-rim distance assessments in non-contrast CT images. The evaluation results demonstrated high accuracy in automated vessel segmentation and distance assessment, showing the feasibility of using the proposed method for risk assessment of vascular injury in THA surgical planning.