Specimens and 3D registration algorithm
Ninety-six cadaver specimens of the lower leg, provided by the Institute of Forensic Medicine, University Zurich and analyzed in a previous study [17], were included without previous trauma, surgery or deformity of the tibia or fibula. Thirty-four male and 12 female donors (missing gender information in two specimens) with an average age of 52 years ±17.7 (range: 21 to 95 years) were included. The average weight was 83.1 ± 16.5 kg (range: 55 to 111 kg) and the average height was 176.2 ± 8.6 cm (range: 154 to 195 cm).
High-resolution computer tomography (CT) data were acquired using a Somatom Definition Flash CT scanner (Siemens®, Erlangen, Germany) with a slice thickness ranging from 0.5 to 0.6 mm. 3D triangular surface models of 96 paired (48 left, 48 right) healthy tibiae and fibulae were generated with manual threshold segmentation and region growing using MIMICS software (MIMICS Medical, Materialise NV, Leuven, Belgium) and imported into the in-house surgical planning software CASPA (Balgrist, Zurich, Switzerland). To approximate the original JL from the mirrored contralateral side, an iterative closest point (ICP) algorithm [1] was used to superimpose the mirrored contralateral model onto the original model, as described in previous studies [17, 25]. A 3D coordinate system was defined according to [5]; z-axis equal directional vector as the anatomical tibia axis defined by an oriented bounding box (OBB) [26], x-axis: lateral, y-axis: anterior (see Fig. 1).
Definition of tibia and fibula segments for contralateral registration
As segment selection and included anatomical structures potentially improve the accuracy to approximate the original model [25], we defined three distinct segments of the lower leg to restore the JL, excluding the potentially deformed tibia plateau. The contralateral lower leg model was mirrored and three anatomical segments were defined (Fig. 1). We included previously described anatomical landmarks as the tibial tuberosity (TT) and fibular tip [18]:
-
Tibia: The segment was defined as 90% of the tibia length.
-
Fibula: The segment included the complete fibula model.
-
Tibial tuberosity (TT) and fibular tip: The segment was defined from 75% to 90% of the tibia length and the complete corresponding proximal fibula segment (see Fig. 1).
The surface registration algorithm to superimpose the mirrored contralateral models onto the original model was repeated for all three defined segments of the tibia and fibula of prespecified lengths, as described above.
Definition of joint line and accuracy of joint line restoration
The JL was defined as the average plane of ten surface registration points on the medial and lateral tibial plateau in a standardized fashion on the rim and the center of the tibial plateau and visualized in Fig. 2. The approximation of the JL from the contralateral side compared to the original JL was measured in mm in direction of the anatomical tibia axis (z-axis) (positive values indicating an elevation of the JL, negative values indicating a distalization) (see Fig. 2B). Additionally, JL error was defined as mean absolute error for each segment.
Measurement of tibia, fibula length and distance of the fibular tip to the joint line
The length of the tibia and fibula model was defined by the OBB [17]. Side-to-side differences are reported as mean absolute differences. The closest distance of the fibular tip to the JL was measured using an automatic surface registration sphere on the highest point of the fibular tip (see Fig. 3).
The JL definition and distance of the fibular tip to the JL measurement were performed by two readers in 20 lower legs to assess accuracy and inter-reader reliability. Intra-reader reliability was not performed due to the highly standardized definition of the surfaces and the mostly automatized measurement procedure.
Statistics
A post-hoc sample size calculation was performed (significance level set: α = 0.05, power level: β = 0.80) to detect a mean JL error of 0.5 mm, assuming a standard deviation of 1 mm. This resulted in a sample size of 36 per group.
Inter-reader reliability was performed using intraclass correlation coefficient (ICC) with a two-way mixed-effect model assuming a single measurement and absolute agreement.
Normal distribution of the data was tested with Shapiro-Wilk’s test and histograms. Data are reported as mean ± standard deviation and range. One-way ANOVA was performed to analyze differences of tibia JL approximations and Kruskal-Wallis for JL error among the three segments. Multiple post-hoc testing was Bonferroni corrected. Differences between gender were analyzed using a non-paired t-test. Gender, height, weight and side-to-side differences of the tibia length, fibula length and fibular tip to JL distance were included in a linear regression model to analyze their influence on JL error and reported as regression coefficient (β; 95% CI). The significance was set < 0.05. Data were analyzed with SPSS version 23 (SPSS Inc., Chicago, IL, USA).