Subject-Specific Spino-Pelvic Models Reliably Measure Spinal Kinematics During Seated Forward Bending in Adult Spinal Deformity

Image-based subject-specific models and simulations are recently being introduced to complement current state-of-the-art mostly static insights of the adult spinal deformity (ASD) pathology and improve the often poor surgical outcomes. Although the accuracy of a recently developed subject-specific modeling and simulation framework has already been quantified, its reliability to perform marker-driven kinematic analyses has not yet been investigated. The aim of this work was to evaluate the reliability of this subject-specific framework to measure spine kinematics in ASD patients, in terms of 1) the overall test-retest repeatability; 2) the inter-operator agreement of spine kinematic estimates; and, 3) the uncertainty of those spine kinematics to operator-dependent parameters of the framework. To evaluate the overall repeatability 1], four ASD subjects and one control subject participated in a test-retest study with a 2-week interval. At both time instances, subject-specific spino-pelvic models were created by one operator to simulate a recorded forward trunk flexion motion. Next, to evaluate inter-operator agreement 2], three trained operators each created a model for three ASD subjects to simulate the same forward trunk flexion motion. Intraclass correlation coefficients (ICC’s) of the range of motion (ROM) of conventional spino-pelvic parameters [lumbar lordosis (LL), sagittal vertical axis (SVA), thoracic kyphosis (TK), pelvic tilt (PT), T1-and T9-spino-pelvic inclination (T1/T9-SPI)] were used to evaluate kinematic reliability 1] and inter-operator agreement 2]. Lastly, a Monte-Carlo probabilistic simulation was used to evaluate the uncertainty of the intervertebral joint kinematics to operator variability in the framework, for three ASD subjects 3]. LL, SVA, and T1/T9-SPI had an excellent test-retest reliability for the ROM, while TK and PT did not. Inter-operator agreement was excellent, with ICC values higher than test-retest reliability. These results indicate that operator-induced uncertainty has a limited impact on kinematic simulations of spine flexion, while test-retest reliability has a much higher variability. The definition of the intervertebral joints in the framework was identified as the most sensitive operator-dependent parameter. Nevertheless, intervertebral joint estimations had small mean 90% confidence intervals (1.04°–1.75°). This work will contribute to understanding the limitations of kinematic simulations in ASD patients, thus leading to a better evaluation of future hypotheses.

Joint kinematics (i.e. relative motion at the joint between two interconnected bodies) were converted to body kinematics (i.e. absolute motion of a body expressed in a fixed ground reference frame) using the API of OpenSim 3.3 (Stanford University, USA). Thereafter, anatomical landmarks on the bodies previously indicated during the mesh-based IV joint definition when creating the model are used (Overbergh et al., 2020). As the location of a landmark on a body is fixed during the motion, its transformation matrix was appended to the body kinematics to obtain the 3D trajectory of each anatomical landmark throughout the trunk flexion motion. Thereafter, six common spino-pelvic parameters in the sagittal plane, i.e. lumbar lordosis (LL), thoracic kyphosis (TK), sagittal vertical axis* (SVA, Figure 1.1), pelvic tilt (PT) and T1 and T9 spino-pelvic inclination (T1-SPI, T9-SPI) were calculated for every time frame (detailed in Table 1.1). The acetabular cavities of pelvis were used to define the sagittal and coronal plane in which the spino-pelvic parameters are expressed (Figure 1.1 Table 1.1.

Calculating range of motion
To determine the range of motion (ROM) of these spino-pelvic parameters over the duration of the trunk flexion motion, the absolute difference between the values at the start (mean of the first three frames) and end (mean of the last three frames) of the motion were used. Due to noise on the kinematic simulation output, a different value would have been obtained if ROM would have been calculated based on the minimal and maximal values, who's bounds would not necessarily encloses the complete motion.

Appendix 2: Training the alignment reconstruction
The operators were allowed to familiarize with the modeling software (Overbergh et al., 2020), guided by a written manual, using the images of the control subject, until they felt confident working with the software. The operator training was concluded with the alignment reconstruction of a plastinated cadaver as in (Overbergh et al., 2020). The plastinated cadaver was rigidly fixated, preventing the spinal alignment to change between upright biplanar images, used for alignment reconstruction, and the supine CT, used as ground truth for evaluating the alignment reconstruction accuracy. All operators performed this test, however, the results of the reconstruction accuracy of operator 1 (O1) were previously published (Overbergh et al., 2020). The same randomized start model was provided to the operators (Figure 2.1.A). After reconstruction, the results were compared to the CT-segmented ground truth (Figure 2.1.B, Table 2.1) for each operator.
The error values for the newly trained operators (O2 and O3) were consistently higher compared to the error value of O1, who developed the method.

Appendix 3: Distributions of the variability on the operator-dependent model parameters
The Shapiro-Wilk test was used to test the normality of the parameters (position and orientation) of the model components (markers, bodies, joints), which indicated non-normal distributions of the variability in virtual marker position (Table 3.1) at the 0.05 significance level (SPSS 26, IBM). Also for the body positions and orientations (Table 3.2) and the joint positions and orientations (Table 3.3), the parameters did not all have a normal distribution. Thereafter, kernel functions were consistently used to estimate all distribution functions from the respective error histograms (Figures 3.1-3.3) (Distribution Fitter, MATLAB, The Mathworks Inc., MA).

Virtual markers
Supplementary  The total standard deviation (SD, blue) at each iteration over all samples and the SD computed over the last 10% of samples (orange), up to that iteration index (x-axis). As the amount of iterations increases, the difference between the total SD and the SD over the last 10% of samples becomes smaller (see also Figure 4.3 for a details on the instance of convergence).   Figure 4.5. Confidence bands (5-95%) for each of the joint angles for subject 2 (S2). All curves have been normalized to their mean value over the length of the motion to allow visualization within the -10° to 10° joint angle range. AR: axial rotation; LF: lateroflexion; FE: flexionextension.