Volumetric Assessment of Blow-Out Fractures With Automated Segmentation Benefits Thinner Computed Tomography Slice Thickness: A Retrospective Case-Control Study

The volume of the orbital vault is an important factor in assessing asymmetry, blow-out fractures, and postoperative outcomes after reconstruction. However, manual segmentation is time-consuming and may depend on the radiologist's experience. The automated segmentation tool speeds up the process, but it may not yield results as similar to manual segmentation. The orbital volume of 34 patients suffering from orbital blow-out fracture requiring surgical intervention and pre- and postoperative computed tomography (CT) data were assessed with an automated segmentation tool (case) and manual segmentation (control). Pre- and postoperative CTs of fractured and intact orbits were assessed (n=136). Slice thickness as a factor affecting the repeatability results of automated segmentation was analysed with the Mann-Whitney U test. Agreement between methods and measurements was assessed using the interclass correlation coefficients (ICC), and the repeatability of each method was tested using intraclass correlation coefficients (ICC). Compared with thicker (2 mm+) slices, 1 mm or less slice thickness showed significantly better repeatability between pre- and postoperative imaging when automated segmentation was used (P=0.028), although overall repeatability was very good in both methods; ICC was 0.83 in manual segmentation and 0.92 in automated segmentation. Correlation between automated and manual segmentation was rather poor in both intact and fractured orbits in pre- and postoperative imaging (ICC: 0.57 and 0.50 in intact orbits; 0.41 and 0.67 in fractured and reconstructed orbits). Automated segmentation is a reliable and repeatable method in volumetric assessment, especially with a slice thickness of 1 mm or less. However, volumetric results do not correlate with manual segmentation.