End ROM for each segment are reported in Figure 3 for each diagnostic movement performed. The data indicate that there was a high degree of symmetry between left/right versions of each diagnostic movement (i.e., end ROM for movements performed to the left or the right were not significantly different in any plane when comparing using an adjusted alpha value of .05/18, or .0028). While none of the values reached statistical significance with this conservative alpha, the range for the 18 segmental level p-values calculated for each of the five left/right diagnostic movements ranged from .0048 to .94. Though not reaching statistical significance, four of the ten p-values that are below .05 are resultant from potential motion discrepancies the L4/L5 segment experienced in the frontal and transverse planes. Composite lumbar ROM was examined for differences as well, with the only meaningful difference in left/right diagnostic movements being in the sagittal plane between flexion left and right movements (p = .0048). All other diagnostic movements showed no statistical differences in composite lumbar range of motion (p = .073 to p = .916).

Due to geometrical properties of the vertebra and as expected, diagnostic movements performed in the transverse or coronal planes (i.e., rotation; lateral bending) also exhibit coupled motions in the coronal or transverse planes, respectively. For example, while performing rotation left and right, subjects exhibit an average of 1.3° of motion in each segment in the coronal plane. Similarly, lateral bending left and right exhibit an average of .6° of motion per segment in the transverse plane. Likewise, flexion exhibits an average of 2.4° of motion in the transverse plane and 2.5° of motion in the coronal plane, indicating coupled motion throughout the movement. A table included in the Supplementary Material includes data as percent coupled values similar to those in a systematic review on lumbar kinematics published by Widmer et al. (2019) for the three primary planes of motion in flexion/extension, lateral bending left and right, and rotation left and right.

In picking up a box positioned on the left as depicted in Figure 4, there was less flexion in the T11/T12 (3.2°) segment when compared to the other FSUs (5.3–10.8°). During ascending motion in this movement, subjects overshot their original starting position in the transverse and coronal planes by an average of 1.2 and .9°, respectively. Additionally, in the transverse plane, the data indicate a semi-periodic motion in the thoracolumbar joint (T11-L2). The peak and trough values for the T11/T12 segment are 3.9 and 1.3°, respectively. The equal time spaces between the (Figure 4A) kinematic visualizations depict the maximum accelerations at the start of the diagnostic movement and the maximum decelerations at the end of the diagnostic movement.

During an Up and Go test as seen in Figure 5, the subjects exhibited periodic motion in the lumbar spine as they walked forwards. The greatest ROM in this periodic motion in the transverse and coronal planes is exhibited in the L4/L5 segment and the smallest is exhibited in the L2/L3 segment.

During the Up and Go test, subjects start with slight flexion in the lower spine as they try to gain momentum to stand up. The lumbar spine then extends into a lordosis as they stand up straight. Since data are shown relative to the subjects’ starting position, the top graph depicts the lordosis of the lower spine once subjects are erect compared to when seated. The data indicate that in standing up, the L3/L4,L2/L3,L1/L2 segments go through a greater ROM in the sagittal plane than the L4/L5, T12/L1, and T11/T12 segments. The L3/L4 segment has an average maximum extension value of 4.8°, the L2/L3 segment has an average maximum extension value of 2.9°, and the L1/L2 segment has an average maximum extension value 1.8°.

Other points of interest in Figure 5 include the offset peaks of the ROM in the transverse and coronal planes. Range of motion in the transverse plane reached local minima and maxima .17 s before the peak in the coronal plane. Subjects also tended to exhibit small amounts (<2°) of lateral bending and axial rotation in the lower spine as they stood up.

Figure 6 reports the average angular velocity of each segment in the primary plane of motion (sagittal for flexion, extension, and box diagnostic movements, frontal for lateral bending, and transverse for axial rotation) travelling towards maximum deflection and returning from maximum deflection. The continuous exercise ball diagnostic movement along with the Up and Go test were removed from this analysis, since they are non-cyclic motions. Hotelling’s T-squared testing was performed to compare the two directions of motion (ascending and descending) and found all diagnostic movements except for exercise ball extension had significant differences (p = 8.07e-14 to p = 0.00033) between the directions of movement for primary plane motion in each diagnostic movement. Segmental level t-tests were performed for all motions that had significant Hotelling’s T-squared test values. A Bonferroni correction factor of six was used to account for testing each segment (α = .0083). Sixty-three of seventy-eight measurements of the segments have significantly different speeds going to maximum deflection versus returning with p-values ranging from 2.33e–12 to .0073. In addition to segmental motion, composite lumbar spine speed was compared as well, with all 14 motions returning low p-values (5.63e–17 to .0176) and reaching significance at α = .05.

Figures 7, 8 report peak angular velocity and acceleration. Peak angular velocities did not follow the same trend of being greater during ascending motion. No meaningful differences in peak descending and ascending velocities were found in any diagnostic movements. While not reaching statistical significance using a Bonferroni correction factor of 18 (α = .0028), there were 11 segments that had p-values below .05 in 8 different diagnostic movements (p = .01 to .048). Most of these segments are a single segment of a single plane in a single diagnostic movement and do not point to an area of further potential interest. However, extension right, box left, and box right all had motion of the L4/L5 FSU in the transverse plane that had p-values below .05 when comparing descending vs. ascending motion.

When comparing symmetrical diagnostic movements, the L4/L5 segment moves at a higher speed when the box movement is performed to the left rather than to the right (p = .0019). While this is the only segment at which the peak velocity difference between right and left diagnostic movements reaches significance at α = .0028, there are six other segments in different planes and diagnostic movements that had p-values below .05 (p = .007 to .042). When examining peak acceleration differences between symmetrical diagnostic movements, there were also 8 other segments in different planes and diagnostic movements that had p-values below .05 (p = .0048 to .049). Low p-values for these diagnostic movements were scattered with the exception of the box left versus box right movements, particularly in the frontal plane. The T11/T12, T12/L1, L1/L2 segments trended to moving faster in the frontal plane during box left movements than during box right (p = .010 to .033). The T11/T12, L1/L2, L2/L3, and L3/L4 segments trended towards accelerating faster in the frontal plane during box left movements than during box right (p = .0048 to .042).

In picking up the box positioned on their left (see Figure 4), the subjects must rotate almost 90° to the reach the box and pick it up. However, the data show that this rotation does largely not occur in the lower spine. The largest angular deflection in a measured FSU is about 2.8° while most of the segments exhibit less than 1.5° of axial rotation. Therefore, the rotation necessary to perform this movement must come from other joints such as the subtalar joint, the hip joint, and/or the upper thoracic spine. Subjects with cLBP may accentuate this or similar trends to minimize motion in the lower back.

When comparing this box movement to flexion (see Supplementary Figure S1), the box movement takes approximately 1 s less to complete and does not exhibit the same plateauing shape in the sagittal plane as uniplanar flexion does. This is likely due to the influence of pelvic motion, commonly described as lumbo-pelvic rhythm during the motion of flexion. The lumbar vertebrae begin to move caudally to rostrally, and as they start to near end ROM, the pelvis becomes the primary driver of motion. Pelvic motion comprises the majority of the motion nearing full trunk flexion, but the spine also has larger end ROM enabled by this lumbo-pelvic rhythm. The extra sagittal end ROM in flexion compared to picking up a box is indicative of the lumbo-pelvic rhythm. Future work could provide further details on kinematics such as further understanding of in-phase versus out-of-phase motion for multiplanar movements or a more detailed look at the L/P ratio broken down by segment (Esola et al., 1996; Kim et al., 2013; Tafazzol et al., 2014; Laird et al., 2018; Ghasemi and Arjmand, 2021). Although it is not the focus of the present work, a closer investigation of additional time-sequenced coupled movement patterns (i.e., spinal rhythms) may provide mechanistic insights into the observed spinal motions.

During the diagnostic movement of picking up a box positioned on the left side, subjects tended to overshoot and laterally bend to the right upon returning to the starting position. This trend was common in many of the movements measured and has been shown in simple flexion tests as well (Sui et al., 2011; Consmuller et al., 2012a). It is possible that the subjects’ trunk right rotators, i.e., left external abdominal obliques, left lumbar multifidi, and right internal abdominal obliques, were activated too much or that their counterparts were not able to properly decelerate the movement, which was amplified by the weight of the box. In the similar diagnostic movement of performing flexion with axial rotation to the left, subjects had similar overshoot in the coronal plane (see Supplementary Material). However, the subjects took an additional .4 s to return to their neutral position after overshooting with the box.

During the Up and Go tests (see Figure 6), the maximum ROM of the periodic movement occurs in the lowest part of the spine. The movement of the pelvis in the coronal plane during the initial steps lead to a large deflection at the bottom of the spine that slowly is compensated as the spine makes up for the movement.

At the 2.5 s mark when the subjects start to walk, it is visible that the largest ROM in the frontal plane occurs at the L4/L5. The L3/L4 segment follows closely. This lateral bending to the right is necessitated by the left hip rising and the lumbar spine counter-side bending. The difference in peak timing during walking suggests that the spine does not rotate in the transverse and coronal planes at the same time, but rather individual muscles such as the rotatores and the erector spinae pull at different times throughout the gait cycle and cause the offset peaks.

We were unable to locate studies that have reported the kinematics of the timed Up and Go test, though a few studies have looked at the kinematics of a sit to stand test (Parkinson et al., 2013; Alqhtani et al., 2015; Christe et al., 2016; Hemming et al., 2018; Pourahmadi et al., 2020). Most of these studies look at total lumbar ROM, and our data fall within the previously reported values.

In contrast to the majority of the diagnostic movements, we observed that several of the box lift and extension diagnostic movements did not exhibit any meaningful difference in descending and ascending average velocities in multiple spinal segments. We hypothesize that the extra weight of the box in the ascending motion slowed subjects down. Trends for flexion and extension disagree with the findings of Consmuller et al. (2012a) in which the average ascending velocity for flexion was greater than the descending velocity, while extension was significantly faster during the descending portion of the movement. It is likely that subject instructions played a role in this. In general subjects were instructed to move at a comfortable speed, however, some subjects in the present work were instructed to move slower during extension movements due to excessive marker dropout during their practice runs.

There is some disagreement in previously published work, with some authors reporting faster average descending velocities [e.g. (McGregor et al., 1995)], and others suggest that is not always true [e.g. (Marras and Wongsam, 1986)]. In the present study, it appears that certain subjects favor moving slower on the ascending motion than the descending motion. Discrepancies in past work may be due to biased samples or small sample sizes, though both the Consmuller et al. (2012b) and the McGregor et al. (1995) studies had over 200 participants.

Average velocities of the entire lower spine in flexion and extension had smaller values as compared with those reported in previous work (Marras and Wongsam, 1986; McGregor et al., 1995; Consmuller et al., 2012a). Consmuller et al. (2012b), Marras and Wongsam (1986), and McGregor et al. (1995) all report larger descending flexion velocities (with standard deviations well within the present work’s data) than the present work reports. Respectively, they report average descending angular velocities of 55, 48, and 30.3 deg/s compared to the present work’s 28.3 deg/s. Average ascending angular velocities also differed with reported respective values of 49, 36, and 51.3 deg/s compared to the present work’s 34 deg/s. Discrepancies in methodology should be also noted, such as the measurement system for the Consmuller study ran further up the thoracic spine, and the measurement systems used in the Marras and McGregor studies included data on the L5/S1 joint. Our measurements for axial rotation and lateral bending are also smaller than the data reported in the McGregor study.

In addition to differences in the spinal region measured, the protocol for calculating average velocity is different in the present work. Measurements in the Consmuller and McGregor studies are restricted to strictly the descending and ascending phases of the motion and ignore the peaks of the motion due to inconsistencies in how long subjects stayed at their maximal ROM. Since the present work also reports maximum angular velocity for each motion, it was decided to include the maximal ROM phase in the analysis to provide data on how these peaks affect average velocities. This inclusion of the maximal ROM phase where subjects slow down and change direction also greatly explains why the present work reports smaller average velocities than previous papers. It also denotes that there are grounds for a more direct comparison between our peak angular velocity data and the average velocity data in these previously published works.

Peak angular velocities recorded appear to be higher than values reported in the Consmuller et al. (2012a) and McGregor et al. (1995) studies mentioned above when accounting for the number of spinal segments measured. In the Marras and Wongsam (1986) study above, subjects were asked to perform prescribed movements at a normal rate as well as at an accelerated rate. Recorded peak velocities of 63 deg/s for descending flexion and 65 deg/s for ascending flexion were well below the accelerated rate (108 deg/s descending, 76 deg/s ascending), but were still higher than the normal rate. While these values are similar, the same methodology discrepancies factor into data differences.

The significantly different descending and ascending average velocities compared to the descending and ascending peak velocities that showed no difference indicate that on average people either tend to take longer to start a diagnostic movement than to finish it or take longer slowing down at maximal deflection than speeding up at maximal deflection. In the latter case, the concentric contraction of the large back and abdominal muscles may be more explosive than the eccentric contraction of the muscles during these trunk movements. Another school of thought is that subjects are more cautious when descending to a new position, and less hesitant when returning to a recently held position. A further psychological factor could be due to watching demonstrations of the research assistants who may have unknown biases towards certain movement patterns.

This was a healthy cohort and symmetrical motion discrepancies were not expected. When looking at the results of the peak angular velocities and accelerations (see Figures 7, 8), it is of no surprise that there are generally no differences in the peak velocities and accelerations when performing a diagnostic movement to the right versus the left for a healthy group of people. However, it is interesting to note the asymmetries in the box movements when comparing descending and ascending velocities. One hypothesis is that the body’s dominant side has stronger muscles and the contractions of the external obliques of the dominant side provide the bulk of the power required to perform a rotation movement. In the current study, handedness was not recorded, however it could be of interest in future work. It is hypothesized that the extra weight of the box makes this a more pronounced difference and is therefore visible compared to the bodyweight only movements.

In comparing our measured values to those found in the literature, velocity and acceleration data are not widely reported at the segmental level in the lumbar spine. When comparing total angular velocities and accelerations however, these data fall within the range of previously reported total lumbar values (Consmuller et al., 2012a; Williams et al., 2014; Vaisy et al., 2015).

One purpose of this study was to assess the feasibility of measuring dynamic segmental kinematics of the lumbar spine during diagnostic movements using a traditional motion capture system. The present work indicates that quantitative estimates of kinematic measurements for each segment can be obtained in all three planes of motion. Accuracy of the measurements was addressed through comparisons with published data gathered through well-established methods of ROM measurement such as the use of bone pins or DFV for segmental motion, and inclinometers for end ROM.

Many systems have been used to measure active end ROM of the spine, but few have reported segmental motion or dynamic kinematics. Cobian et al. (2013) measured lumbar ROM and rates of excursion in subjects performing tasks of daily living using a motion capture system. They showed total lumbar excursion rates for daily movements to generally be well below a healthy person’s maximal excursion rates when the subjects performed simple flexion, lateral bending, and axial rotation movements (Cobian et al., 2013). Cook et al. (2015) measured cadaveric lumbar FSU ROM in an ex vivo environment using a motion capture system. They reported detailed segmental ROM for the lumbar region when testing flexion/extension, lateral bending, and axial rotation and compare their results with previous ex vivo and in vivo findings (Cook et al., 2015). However, their findings, along with other cadaveric studies (Cholewicki et al., 1996; Zirbel et al., 2013; Stolworthy et al., 2014), reported no time dependence in the motions, and do not capture the same details that measurements on living subjects would. Rozumalski et al. (2008) measured segmental ROM data during gait as well as flexion/extension, lateral bending, and axial rotation on a cohort of live subjects using bone pins and a motion capture system (Rozumalski et al., 2008). Our data compare favorably to the gold standard of bone pin measurements as performed by Rozumalski et al. Consmuller et al. (2011) and Suter et al. (2019) measured lordosis angles in the spine using a strain gauge system. Consmuller et al. measured segmental end ROM, and Suter et al. took continuous data on the average lordosis of the lumbar spine with both a motion capture system and the strain gauge system, but these measurements only examined sagittal plane motion (Consmuller et al., 2012b; Suter et al., 2020).

Our end ROM flexion and extension values gave similar results (5.1% larger than flexion and extension values that were measured using bone pins, and segmental end ROM measurements ranged from 2.4% to 24.1% different) (Rozumalski et al., 2008). Flexion and extension values from our trials also came within 10% of previous studies that used inclinometers or motion capture to measure end ROM in the lower spine (Cobian et al., 2013; Kolber et al., 2013). Values for lateral bending and axial rotation were less consistent with previous data, with total end ROM values ranging from being 40.0%–66.9% smaller.

Segmental end ROM ranged from 24.1% greater to 101.1% smaller than previously reported values (Rozumalski et al., 2008). Potential confounding factors besides skin artifacts include differences in instructions given to the subjects performing the diagnostic movements and reference points for the measurements. With the focus on functional movements, our subjects were to move as far as was comfortably possible, but not to exert themselves to go as far as possible. In addition, for movements such as axial rotation, subjects were instructed to keep their hips facing anteriorly which reduces FSU end ROM (Moreside et al., 2013). The studies with the compared ranges do not always give details on the instruction subjects received for these movements. Some of the reported differences may also be related to distinct subject demographics. For example, a study performed by Rozumalski et al. studied 10 healthy subjects but does not include data on the age of the subjects. It is possible that they had a different age demographic than the average age of 25 in this study.

All skin-mounted marker measurements are susceptible to influence from the locational placement of the markers. Many studies have quantified this effect in various parts of the body (Della Croce et al., 2005; Gedet et al., 2007; McFadden et al., 2020; Severijns et al., 2021). Severijns et al. (2021) focused on the effect of marker placement in measuring spinal kinematics in subjects with adult spinal deformities. In their control population, they found that mediolateral palpation errors had significant correlations to BMI and soft tissue thickness. Inferosuperior palpation errors did not show the same correlations.

Numerical simulation was undertaken to examine the effects of marker misplacement on segmental kinematics. The L4C marker was shifted 5 mm in the -Z direction, 5 mm in the +Y direction, followed by a combination of 5 mm in both the Y and -Z directions.

Simulation for the superior segment affected by a central marker (see Figure 9) shows that the method is relatively insensitive to marker misplacement during Flexion and Axial Rotation but exhibits a higher sensitivity to marker misplacement (particularly in the Z-axis) during Lateral Bending. Central marker sensitivity was examined since the lateral markers were measured off the palpated central markers, and it is expected the greatest source of variation in both intra- and inter-operator reliability would originate from variations due to skill in palpating the spinous processes. A similar figure for the effects on the inferior segment can be found in the Supplementary Material.

The high-density marker set used in the present work provides an advantage in that calculation of segmental lumbar angles is possible (as opposed to a gross overall measure of total lumbar angle). However, the high-density marker set may magnify the local systematic error measures of those calculations in comparison to methods that examine total lumbar kinematics.

Soft tissue artifacts are an additional source of error. These artifacts include both motion artifacts caused by inertial forces as well as artifacts caused by the skin and underlying bones moving at different rates. The soft connective tissues such as the supraspinal ligament, subcutaneous layer (including adipose tissue), and cutaneous membrane that compose the low-stiffness connection between the spinous processes and the skin surface where the markers are located introduces a region for greater error that grows as the overall thickness of the soft tissue layers grow (Sistrom et al., 1993; Kawchuk et al., 2011). Soft tissue artifacts are both task dependent and subject specific (Leardini et al., 2005; Zemp et al., 2014). Zemp et al., and Morl et al., both measured marker shift on skin in relation to the spinous processes of lumbar vertebrae using vertical MRI machines (Morl and Blickhan, 2006; Zemp et al., 2014). Zemp et al. found an average skin shift of 7.1 mm for L1 through L5 in the inferior-superior direction in flexion with a maximum shift of 27.4 mm while Morl et al. found a median skin shift over the L3 and L4 spinous processes of .86 mm with a maximum shift of −9.86 mm in three different positions.

Zemp et al. also reports lumbar curvature as measured by markers and by MRI for L1-L5 and show average discrepancies of 17.1–21.8° depending on position (Zemp et al., 2014). Similarly reported discrepancies in the thoracic kyphosis (T3-T11) ranged from −8 to 4.7°. Their work states that foam tubes were attached to subjects’ backs to protect the markers while sitting against a backrest. However, the attachment method is not reported and may have affected the skin stretch in different postures (Cimino et al., 2018). In all, these works indicate that motion trends may be captured by optical motion markers, but they call for caution and awareness of the shortcomings in interpreting the results.

To estimate effects of STA’s on the present data, numeric estimations were made ranging from the 0.86 mm median to the 7.1 mm average inferior-superior STA’s cited above for flexion. Ramped deflections in both the positive and negative Y were applied to the L4C marker starting 1 s before and ending 1 s after the peak of the motion as found by MATLAB’s findpeaks function. Figure 10 depicts the results of this simulation. While this simulation does not account for any errors between the markers and the vertebrae in the initial orientation, it quantifies potential differences in true vertebral body motion compared to motion as measured by the markers. It is anticipated that similar changes would happen in other planes in other motions, but few spinal STA quantification data are available outside of flexion-extension.

An important aspect of interpreting these data, especially for periodic motions, is understanding the normalization of the data. Picking the landmark about which to normalize the data can impact the results. Average range of motion at the normalization points are higher, and other areas tend to have lower values for the segmental range of motion. This is due to greater variations in subjects’ data the further away from the point of normalization. It is particularly notable in the periodic exercises Up and Go, and Exercise Ball Continuous. All other diagnostic movements were non-periodic and show less variation based on normalization point with uniplanar motions showing the least variation, and multiplanar motions showing slightly more variation based on the chosen normalization point.

Both a 6 Hz low-pass Butterworth filter and a smaller sized,51-frame, Savitsky-Golay filter were considered before settling on the final Savitsky-Golay filter used in the present work (see Figures 11, 12). As can be observed in the figures, the process of smoothing the coordinates affects the velocity and acceleration calculations. Data obtained using the final choice of filter compared favorably with peak velocities and accelerations found in previously published studies (Ma et al., 2009; Tsang et al., 2017).