CT-Based Three-Dimensional Volumetric Analysis of Posterior and Lateral Malleolar Fragments in SER-Type Trimalleolar...

Posterior malleolar fractures are commonly assessed using two-dimensional measurements and morphology-based classification systems. However, ankle fracture morphology is inherently three-dimensional, and the reproducibility of CT-based volumetric segmentation for malleolar fracture fragments has not been sufficiently established.

This study aimed to evaluate the relationship between lateral and posterior malleolar fragment volumes in homogeneous supination-external rotation (SER)-type trimalleolar ankle fractures and to assess the intraobserver and interobserver reproducibility of a manual CT-based three-dimensional volumetric segmentation workflow.

This retrospective musculoskeletal imaging study included 71 patients with SER-type trimalleolar ankle fractures who underwent preoperative computed tomography (CT). Posterior and lateral malleolar fracture fragments were segmented on thin-slice axial CT images using a standardized manual contour-based slice-by-slice workflow. Fragment volumes were calculated using dedicated volumetric imaging software. The association between lateral and posterior malleolar fragment volumes was assessed using Spearman correlation and multivariable linear regression analyses. Measurement reproducibility was evaluated using intraclass correlation coefficients for absolute agreement [ICC(A,1)] and Bland-Altman analyses.

The median lateral and posterior malleolar fragment volumes were 8.63 cm³ (interquartile range [IQR], 7.18-10.71) and 2.64 cm³ (IQR, 1.88-4.24), respectively. A weak but statistically significant positive correlation was observed between lateral and posterior malleolar fragment volumes (Spearman rho = 0.313, p = 0.008). In multivariable linear regression analysis, lateral malleolar fragment volume remained independently associated with posterior malleolar fragment volume after adjustment for age, sex, and body mass index (B = 0.316, standardized β = 0.39, p = 0.002). Intraobserver and interobserver reproducibility were excellent for all volumetric measurements, with ICC(A,1) values ranging from 0.996 to 0.999. Bland-Altman analyses demonstrated low mean bias and narrow limits of agreement across all comparisons. Geometric agreement was also excellent, with Dice similarity coefficient values ranging from 0.93 to 0.96 across intraobserver and interobserver segmentation comparisons.