Gait is a dynamic activity that can be described as synchronized, repetitive, multiplanar rotational movements of body segments around joints. Detailed analysis of limb movement during locomotion, referred to as kinematic analysis, provides a better understanding of limb movement in terms of lengths, angles, velocities, and accelerations of body segments. In contrast to physical examinations that involve passive movements and manipulations of subjects, kinematic analysis involves subjects that exercise voluntary control over their musculoskeletal and nervous systems and therefore provides more relevant information regarding limb function. Kinematic gait analysis has been used to recognize and characterize clinically normal gait patterns and identify abnormal gait patterns related to pathologic conditions in humans and other animals.1–3
Because every body movement including gait is 3-D, 3-D kinematics has been accepted as a standard of gait analysis.4 Modern 3-D analysis systems identify the location of anatomic skeletal segments in 3-D space via mathematic algorithms of computer programs. The final images generated by those systems are similar in depth to those perceived by human stereoscopic vision. Limitations of 3-D kinematic analysis systems include the high cost of specialized equipment and computer software. Although the development of advanced, user-friendly hardware and software has simplified data collection substantially, proper operation of the system and interpretation of the collected data still require knowledge of and experience with these sophisticated systems. In addition, the use of 3-D kinematic analysis in veterinary medicine is limited by the lack of basic data (including canine body measurements) required for proper analysis, the difficulty locating accurate centers of joints in 3-D space, and the substantial conformational variability within and among breeds of dogs.
Two-dimensional systems for gait analysis are simpler, easier to use, and more economical than 3-D systems. The simplest configuration may require just a conventional video camera and some analysis software. In addition, 2-D and 3-D analysis share a main principle (video-based photogrammetry), which is the basis of current theories and techniques in gait analysis.4 The main limitation of 2-D analysis is that it assumes that all motion takes place in 1 plane (calibration plane). Therefore, when a motion is taking place in a plane far from the calibration plane or a plane not parallel to the calibration plane, the recorded motion may be exaggerated or distorted, and the subsequent analysis may not be accurate.
Gait analysis can be used to aid in diagnosis and to determine treatment outcomes in clinical situations. However, establishing a gait laboratory with modern 3-D systems requires a substantial investment. Proper analysis at various gait velocities requires at least 10 m of open space for a walkway and additional space for the expensive (> $100,000) equipment. In addition, a gait laboratory should be close to the clinic for easy access of animals with abnormal limb function. Given the aforementioned limitations of 3-D analysis in veterinary medicine, it may be difficult to justify investment of such resources to establish a 3-D system for routine clinical situations. However, it may be feasible to start with 2-D analysis of gait by use of a video camera for clinical purposes, such as general description of limb movement in a craniocaudal direction instead of a precise joint motion with 6 degrees of freedom.
The primary purpose of the study reported here was to evaluate a simple 2-D system (a digital video camera and free software) for description of canine hind limb motion in the sagittal plane during walking. Because canine limbs are located vertically and most movement of these limbs during walking is in the sagittal plane, the 2-D system may characterize hind limb movement adequately without substantial out-of-phase error for a fraction of the cost of a 3-D system. We therefore tested the hypotheses that kinematic data of the sagittal motion of canine hind limbs during walking obtained with a 2-D system would correlate well with similar data obtained with a 3-D system and that the data obtained via the 2-D system would be repeatable.
Materials and Methods
Animals—Six adult dogs with no evidence of lameness were used in the study (age range, 5 to 10 years; mean body weight, 32.9 kg). Two dogs were male; 4 were female. Breeds represented included Labrador Retriever (n = 4), Golden Retriever (1), and Greyhound (1). The study protocol was approved by the Purdue Animal Care and Use Committee.
Data acquisition—Each marker set was constructed from polystyrene ballsa covered with reflective tapeb (diameter, 1.9 cm) and IREDsc (diode diameter, < 0.1 cm; collar diameter, 0.7 cm). A flat IRED was glued onto the vertex of each reflective marker and positioned vertically so the centers of both markers coincided when viewed from the side. Marker sets were attached to the skin on defined anatomic sites2 on the left hind limb of each dog by the same investigator as follows: the top of the thoracolumbar junction, the iliac crest, the greater trochanter of the femur, the femorotibial joint between the lateral epicondyle of the femur and the fibular head, the lateral malleolus of the distal aspect of the tibia, and the distolateral aspect of the fifth metatarsus. The IRED cablesd were wrapped around the hind limb and caudal abdomen with hook-and-loop fasteners and elastic tapes so the gait of the dog and reflection from the markers would not be impeded. Cables were then connected to a strober held by a handler. All dogs walked comfortably with markers and elastic tapes on the body.
Dogs were led on a leash by a handler and walked at a comfortable velocity in a straight line on a force platee mounted in a 10-m wooden walkway. Ten trials were recorded for each dog, and the first 5 valid trials were used for data analysis. A valid trial included a straight, forward walk without stopping, hesitating, or running; no overt head movement during walking; and a left hind limb strike on the force plate. In addition, all marker sets had to be recognized by cameras throughout the trial. As each dog walked on the walkway, kinematic data of the left hind limb were simultaneously collected via the 2-D video and 3-D optoelectric systems.
2-D video system—A digital video cameraf (60 Hz; shutter speed of 0.004 seconds) was positioned 3.5 m from the force plate. The camera was located along the left side of the walkway, and a 600-W halogen lightg was positioned just above the camera. A testing area (reference area) 2 m in length was established, and the camera was centered and focused at the midpoint of that area. The end of the force plate was set at the midpoint of the testing area, which also served as the origin of the global reference system for the 3-D system. Horizontal linear calibration of the video system was performed at the beginning of the recording and consisted of manual digitization of 2 horizontal points of 2 reflective markers, which were set 2 m apart. Once a gait trial had started, the video camera recorded all trials of a dog without stopping or pausing.
3-D optoelectric system—Ground reaction force and kinematic data were collected at 180 Hz for 4 s/trial by means of an optoelectric system.h The 3-D system was calibrated in the testing volume with the origin of the global reference system located at the end of the force plate. A calibrated space that was 2.5 m long, 1.5 m wide, and 1.5 m high was recorded to enable tracking of markers throughout the gait cycle of the left hind limb. Two sets of 3 cameras (total of 6 cameras) were arranged in a semicircle on the left side of the dog and the runway to record walking. For each trial, the 3-D optoelectric system was triggered manually and stopped automatically after 4 seconds.
Data processing—Video recordings for the 2-D analysis were processed by use of designated software.i Trajectories of all reflective markers of each valid trial were manually digitized by use of that software. For the 3-D analysis, the coordinates of the markers were processed via the software of the 3-D optoelectric system.
Displacement data of markers, represented as x (craniocaudal) and y (vertical) coordinates (2-D) or x, y, and z (mediolateral) coordinates (3-D), were acquired, and the data from all recorded frames were subsequently exported to a worksheetj (2-D 60-Hz and 3-D 180-Hz data sets, respectively). To determine the effect of a reduction in the rate at which data were acquired from 180 Hz to 60 Hz, every third frame of the 3-D data set was saved and stored in a separate file (3-D 60-Hz data set) via custom-designed software.k For the effect of spatial difference, 3-D data were condensed to only x and y coordinates (3-D sagittal 180-Hz data set). In addition, the rate of the 3-D sagittal 180-Hz data set was reduced to 60 Hz, and the data were saved as a separate file (3-D sagittal 60-Hz data set). Thus, 5 sets of data (2-D 60 Hz, 3-D 180 Hz, 3-D 60 Hz, 3-D sagittal 180 Hz, and 3-D sagittal 60 Hz) for each dog were created and each set of data was filtered with a fourth-order Butterworth low-pass filter, with a cutoff frequency of 8 Hz (selected by residual analysis of several trials) to attenuate noise in the signal.5
Dynamic joint angles in the sagittal plane were determined for hip, stifle, and tarsal joints of the left hind limb. Joint angles were automatically calculated during data acquisition by use of the following equation:
Each trial included 1 complete gait cycle of the left hind limb, which started when the left hind limb contacted the force plate (beginning of stance phase) and finished when the left hind limb contacted the walkway during the subsequent gait cycle. The initiation of the stance phase of the analyzed stride was identified by visual estimation of the point of contact of the left foot with the force plate (2-D) or by the first positive force signal in the vertical force curve (3-D). For the 2-D system, termination of the analyzed stride was determined by visual estimation of the point at which the left hind limb subsequently contacted the walkway. For the 3-D system, the exact frame of stride termination was subsequently identified as the frame with the biggest braking velocity of the x coordinate of the metatarsal marker. This method was based on the kinematic pattern when foot landing was exactly identified by the force plate data. In the reduced 3-D data sets, the start frame was the same as that in the 3-D 180-Hz data set, but the end frame was the same or 1 or 2 frames before the final frame of the 3-D 180-Hz data set. Thus, complete and reduced 3-D data sets always had the same start frame but did not always have the same end frame.
To compare motions during a complete gait cycle, 2-D and 3-D data regarding the sagittal motion of hip, stifle, and tarsal joints during a complete gait cycle were normalized. The first and last frames used were considered 0% and 100% of the gait cycle, respectively.
The following values were determined for each of the 5 data sets (2-D 60 Hz, 3-D 180 Hz, 3-D sagittal 180 Hz, 3-D 60 Hz, and 3-D sagittal 60 Hz) for the individual dogs and for the group after merging the data on individual dogs: mean forward velocity, stride length, gait cycle duration, stance phase duration, duty factor (duration of stance phase/duration of gait cycle), mean joint angle during the gait cycle, peak extension and flexion angle of joints during the gait cycle, angular excursion (peak extension angle – peak flexion angle) of joints during the gait cycle, and CV (%) of each joint angle data. To determine the difference between the 2-D and 3-D systems, the mean absolute difference at each normalized time point was calculated.
Statistical analysis—To examine the agreement between data obtained by use of the 2-D and 3-D kinematic analysis systems, a Bland-Altman test was performed to compare mean differences and respective SDs for the 6 dogs.7 Differences ± (1.96 × SD) for each joint are reported as the limits of agreement, which means that 95% of the differences between the 2-D and 3-D joint angle data were located within these ranges.
To compare the 5 data sets, the mean of 5 trials for each variable was calculated for each dog, and a 1-way ANOVA was performed to identify significant differences among data sets. All analyses were performed with standard statistical software.l Differences were considered significant at P < 0.05.
Results
Temporospatial variables—Mean forward velocity of canine left hind limbs for all 3-D data sets was 1.1 m/s; the velocity for the 2-D data set was 1.3 m/s (Table 1). The duty factor of canine hind limbs for all 3-D data sets was 0.5, and that for the 2-D data set was 0.6, which indicated that all dogs walked, although some dogs walked faster than others. Mean forward velocity, stride length, and duty factor of hind limbs as measured by the 2-D system were higher than those estimated by use of each of the 3-D data sets. Although durations of the gait cycle for each of the 5 data sets were the same, the stance phase as measured by the 2-D system was longer than that determined by use of the other data sets. However, this difference was not significant (P = 0.08). Differences among the four 3-D data sets with respect to values of respective temporospatial variables were not significant.
Mean ± SD temporospatial characteristics of left hind limbs of 6 walking dogs (5 trials/dog) evaluated by use of 2-D video camera and 3-D optoelectric systems.*
| System | Mean forward velocity (m/s) | Stride length (m) | Gait duration (s) | Stance duration (s) | Duty factor† |
|---|---|---|---|---|---|
| 2-D 60 Hz | 1.3 ± 0.3‡ | 1.0 ± 0.1‡ | 0.8 ± 0.1 | 0.5 ± 0.1 | 0.6 ± 0.0‡ |
| 3-D 180 Hz | 1.1 ± 0.1 | 0.8 ± 0.1 | 0.8 ± 0.1 | 0.4 ± 0.1 | 0.5 ± 0.0 |
| 3-D sagittal 180 Hz | 1.1 ± 0.1 | 0.8 ± 0.1 | 0.8 ± 0.1 | 0.4 ± 0.1 | 0.5 ± 0.0 |
| 3-D 60 Hz | 1.1 ± 0.1 | 0.8 ± 0.1 | 0.8 ± 0.1 | 0.4 ± 0.1 | 0.5 ± 0.0 |
| 3-D sagittal 60 Hz | 1.1 ± 0.1 | 0.8 ± 0.1 | 0.8 ± 0.1 | 0.4 ± 0.1 | 0.5 ± 0.0 |
Data recorded by the 3-D system were used to create 4 data sets: original 3-D data recorded at a rate of 180 Hz (3-D 180 Hz), reduced data set in which only the x and y coordinates from the original 3-D data set were stored (3-D sagittal 180 Hz), reduced data set in which every third frame of the original data set was stored (3-D 60 Hz), and reduced data set in which every third frame and only the x and y coordinates of the original data set were stored (3-D sagittal 60 Hz).
Duty factor = Duration of stance phase/duration of gait cycle.
Significantly (P < 0.05) different from other values in the same column.
Joint angles—Mean joint angles, peak extension angles, and peak flexion angles of hip, stifle, and tarsal joints differed among dogs (Table 2). However, when the data were regrouped into 5 data sets with different planes and rates of data acquisition, no significant differences in joint angles among data sets were detected regardless of plane or rate. Angular excursion of the hip joint for the 2-D data set was greater than angular excursions for all 3-D data sets, but those of stifle and tarsal joints were similar among data sets (Table 3).
Mean angles and ranges of hind limb joints of 6 walking dogs (5 trials/dog) evaluated by use of 2-D video camera and 3-D optoelectnc systems.*
| System | Joint | Dog 1 | Dog 2 | Dog 3 | Dog 4 | Dog | Dog 6 |
|---|---|---|---|---|---|---|---|
| 2-D 60 Hz | Hip | 129.5 | 119.9 | 128.2 | 125.0 | 134.8 | 120.5 |
| (150.7–114.9) | (138.2–102.3) | (148.8–110.8) | (147.1–107.7) | (149.7–122.3) | (138.1–108.0) | ||
| Stifle | 147.2 | 135.7 | 146.6 | 143.8 | 131.0 | 146.2 | |
| (171.1–120.6) | (155.0–104.9) | (161.1–120.9) | (155.7–122.4) | (152.7–97.2) | (168.7–105.8) | ||
| Tarsal | 152.1 | 140.9 | 154.2 | 150.7 | 132.9 | 150.1 | |
| (167.9–139.6) | (159.1–118.4) | (175.1–136.2) | (160.1–139.4) | (153.8–99.4) | (167.0–120.7) | ||
| 3-D 180 Hz | Hip | 129.6 | 119.0 | 128.4 | 125.6 | 132.0 | 119.1 |
| (147.7–117.8) | (135.7–101.3) | (147.5–112.9) | (144.4–109.9) | (147.3–119.7) | (135.8–107.6) | ||
| Stifle | 146.8 | 130.8 | 138.9 | 137.0 | 126.0 | 140.3 | |
| (171.0–121.8) | (150.6–103.2) | (154.4–117.3) | (150.9–116.6) | (148.8–95.9) | (164.6–102.8) | ||
| Tarsal | 153.6 | 139.6 | 150.2 | 147.4 | 133.1 | 147.3 | |
| (169.8–145.2) | (162.0–118.3) | (172.5–133.0) | (155.7–135.5) | (152.4–105.3) | (166.2–122.9) | ||
| 3-D sagittal 180 Hz | Hip | 131.2 | 119.5 | 128.9 | 125.7 | 135.8 | 120.8 |
| (150.4–118.8) | (135.9–102.7) | (147.4–114.2) | (144.7–109.3) | (149.7–125.7) | (138.0–109.3) | ||
| Stifle | 148.4 | 132.7 | 143.6 | 139.9 | 129.2 | 142.0 | |
| (173.7–122.2) | (152.4–103.3) | (159.0–118.9) | (153.2–119.5) | (150.2–95.2) | (166.8–103.2) | ||
| Tarsal | 156.4 | 140.5 | 151.7 | 148.9 | 135.1 | 150.4 | |
| (173.3–146.8) | (163.0–118.3) | (173.3–134.0) | (158.0–137.6) | (155.4–104.9) | (167.9–124.2) | ||
| 3-D 60 Hz | Hip | 129.8 | 119.0 | 128.1 | 125.5 | 131.8 | 119.1 |
| (147.6–117.8) | (135.7–101.3) | (147.3–113.6) | (144.4–109.9) | (147.2–119.7) | (135.8–107.6) | ||
| Stifle | 146.5 | 130.9 | 139.2 | 137.1 | 126.3 | 140.4 | |
| (170.6–121.9) | (105.6–103.3) | (154.3–117.6) | (150.8–116.6) | (148.7–95.5) | (164.6–102.7) | ||
| Tarsal | 154.0 | 139.7 | 150.0 | 147.4 | 133.2 | 147.3 | |
| (169.6–145.2) | (162.0–118.3) | (171.9–133.0) | (155.7–135.6) | (152.3–105.3) | (166.1–122.9) | ||
| 3-D sagittal 60 Hz | Hip | 131.2 | 119.4 | 128.8 | 125.2 | 135.7 | 120.8 |
| (150.2–118.9) | (135.9–102.7) | (147.3–114.2) | (144.8–109.3) | (149.7–125.7) | (138.0–109.2) | ||
| Stifle | 147.0 | 132.8 | 143.7 | 140.4 | 129.4 | 142.0 | |
| (171.2–122.1) | (152.4–103.3) | (159.0–119.0) | (153.3–119.5) | (150.0–95.2) | (166.7–103.2) | ||
| Tarsal | 156.4 | 140.6 | 151.7 | 149.1 | 135.2 | 150.4 | |
| (172.9–146.8) | (162.9–118.4) | (173.4–134.0) | (157.9–137.6) | (155.4–104.9) | (167.8–124.3) |
Values reported are degrees. Numbers in parentheses represent the range, which extends from peak extension to peak flexion.
See Table 1 for key
Mean ± SD joint angular excursions during 1 gait cycle of hind limb joints of 6 walking dogs (5 trials/dog) evaluated by use of 2-D video camera and 3-D optoelectric systems.*
| System | Hip joint | Stifle joint | Tarsal joint |
|---|---|---|---|
| 2-D 60 Hz | 34.5 ± 5.0‡ | 48.8 ± 10.2 | 38.3 ± 11.7 |
| 3-D 180 Hz | 31.4 ± 3.3 | 47.1 ± 9.9 | 36.3 ± 10.7 |
| 3-D sagittal 180 Hz | 31.0 ± 4.2 | 48.9 ± 10.2 | 37.5 ± 11.1 |
| 3-D 60 Hz | 31.4 ± 3.4 | 47.0 ± 9.9 | 36.2 ± 10.7 |
| 3-D sagittal 60 Hz | 31.0 ± 4.2 | 48.4 ± 10.2 | 37.4 ± 11.2 |
Comparison of normalized joint motion between 3-D 180 Hz and 2-D 60 Hz—Overall mean normalized joint angle of hip, stifle, and tarsal joints (5 trials from each of 6 dogs/joint) were plotted to compare differences between the 3-D 180-Hz and 2-D 60-Hz data sets (Figure 1). Graphs of the hind limb joint angles measured by the 3-D and 2-D systems resembled a reported3 pattern of hind limb joint angles during walking. The trajectories of hind limb joint angles measured by the 3-D and 2-D systems were similar throughout the entire gait cycle. However, the time point at which the angle of peak flexion and extension was detected during walking via the 2-D system was delayed, compared with the point at which the angle was detected via the 3-D system. With the 3-D system, peak flexion and peak extension of the hip joint appeared to be at 53% and 91% of the gait cycle; with the 2-D system, this appeared to be at 58% and 100% of the gait cycle. A similar pattern was detected for the stifle and tarsal joints. For the stifle joint, peak flexion appeared to be at 96% of the gait cycle via the 3-D system and at 100% of the gait cycle via the 2-D system; respective values for peak extension were 72% and 76%. For the tarsal joint, peak flexion appeared to be at 59% of the gait cycle via the 3-D system and at 62% of the gait cycle via the 2-D system; respective values for peak extension were 77% and 82%, but the magnitudes of delay were smaller than in the hip joint. The excursions of each joint during the gait cycle detected via the 2-D system were larger than those detected via the 3-D system, with differences of 4.4° at the hip joint, 3.7° at the stifle joint, and 3.8° at the tarsal joint.
Overall mean joint angle of hip joints (top row), stifle joints (middle row), and tarsal joints (bottom row) in left hind limbs of 6 dogs evaluated 5 times during 1 walking gait cycle by use of a 3-D optoelectric system (red line; left and middle columns) and a 2-D video camera system (black line; middle and right columns). Error bars in graphs in left and right columns represent SD.
Citation: American Journal of Veterinary Research 69, 9; 10.2460/ajvr.69.9.1116
Repeatability of data—All data on joint angles obtained during the 5 trials of each dog were highly repeatable, as indicated by a low CV (Table 4). The lowest CV overall was 0.3% (mean joint angle of hip and tarsal joints in 1 dog), and the highest CV was 6.2% (peak flexion of the tarsal joint in another dog). With regard to types of joints, the highest CVs were obtained for stifle joints, followed by tarsal and hip joints. With regard to types of angles, the highest CVs were obtained for peak flexion angles, followed by peak extension angles and mean joint angles. The CVs of the joint angles of the 2-D and 3-D data sets were not significantly different.
Ranges of CV of angular measurements of the joints of 6 walking dogs (5 trials/dog) evaluated by use of 2-D video camera and 3-D optoelectric systems.*
| System | Hip joint | Stifle joint | Tarsal joint | ||||||
|---|---|---|---|---|---|---|---|---|---|
| Mean joint angle | pExt | pFlex | Mean joint angle | pExt | pFlex | Mean joint angle | pExt | pFlex | |
| 2-D 60 Hz | 0.3–1.7 | 0.6–1.7 | 0.6–3.0 | 0.3–2.0 | 0.6–2.4 | 0.8–3.6 | 0.6–2.5 | 1.0–1.7 | 0.7–6.2 |
| 3-D 180 Hz | 0.6–1.3 | 0.5–1.4 | 0.6–2.3 | 0.4–1.8 | 0.5–1.6 | 1.0–4.8 | 0.6–1.7 | 0.4–1.5 | 0.8–3.8 |
| 3-D sagittal 180 Hz | 0.7–1.9 | 0.6–2.2 | 0.9–3.5 | 0.4–1.8 | 0.5–1.8 | 1.1–4.4 | 0.5–1.6 | 0.6–1.7 | 0.8–2.9 |
| 3-D 60 Hz | 0.5–1.5 | 0.5–1.4 | 0.6–2.3 | 0.5–1.8 | 0.5–1.6 | 0.9–4.8 | 0.4–1.7 | 0.4–1.8 | 0.8–3.8 |
| 3-D sagittal 60 Hz | 0.7–1.9 | 0.6–2.2 | 0.9–3.5 | 0.4–1.5 | 0.5–1.9 | 1.1–4.4 | 0.5–1.6 | 0.4–1.7 | 0.6–2.9 |
Values reported are percentages. pExt = Peak extension angle. pFlex = Peak flexion angle.
See Table 1 for remainder of key.
Accuracy and agreement of data—Mean absolute differences at each time point (percentage duration of gait cycle) from the normalized angle data indicated that the differences between the 3-D 180-Hz and 2-D 60-Hz data sets were significantly larger than those among the 3-D 180-Hz data set and the other 3-D data sets for all joints (Table 5). When the 3-D data sets were compared, the difference between 3-D 180-Hz and 3-D sagittal 60-Hz (same plane and rate of data acquisition as 2-D 60-Hz) data sets was the largest. The largest differences in all data sets compared were for stifle joints, followed by tarsal and hip joints. When 2-D 60-Hz and 3-D sagittal 60-Hz data sets were compared, the differences between all joints were larger than similar comparisons among the other 3-D data sets. However, differences between the 2-D 60-Hz and 3-D sagittal 60-Hz data sets were smaller than were differences between the 3-D 180-Hz and 2-D 60-Hz data sets.
Mean absolute differences and limits of agreement of angular measurements among data sets for 1 normalized gait cycle of 6 walking dogs (5 trials/dog) by use of 2-D video camera and 3-D optoelectric systems.*
| Data set comparison | Mean absolute difference (°) | Limit of agreement (°) | ||||
|---|---|---|---|---|---|---|
| Hip joint | Stifle joint | Tarsal joint | Hip joint | Stifle joint | Tarsal joint | |
| (3-D 180 Hz) vs (2-D 60 Hz) | 3.7a | 8.1a | 5.3a | 2.5 to 4.8 | 5.4 to 10.8 | 2.0 to 8.5 |
| (3-D 180 Hz) vs (3-D sagittal 180 Hz) | 0.9b | 13b,A | 1.0b,A | −0.9 to 2.7 | −0.9 to 3.5 | −0.6 to 2.7 |
| (3-D 180 Hz) vs (3-D 60 Hz) | 2.0 | 3.3b,A | 2.3b,A | −0.5 to 4.5 | 0.4 to 6.3 | 0.6 to 4.0 |
| (3-D 180 Hz) vs (3-D sagittal 60 Hz) | 2.2 | 3.5b,A | 24b,A | −0.2 to 4.5 | 0.9 to 6.0 | 1.2 to 3.6 |
| (2-D 60 Hz) vs (3-D sagittal 60 Hz) | 2.3 | 5.2b | 4.4B | 1.6 to 3.1 | 1.7 to 8.7 | 1.6 to 7.1 |
Within the same column, values with different lowercase superscript letters differ significantly (P < 0.05).
Within the same column, values with different uppercase superscript letters differ significantly (P < 0.05).
See Table 1 for remainder of key.
Discussion
The 2-D video camera approach to analyzing the kinematics of canine hind limb joints was repeatable and agreed well with the 3-D optoelectric method of kinematic data acquisition. Joint angles and angular excursions of the 2-D data set were similar to those of the 3-D data sets. However, linear measurements of the 2-D system were significantly different from those of the 3-D system. In addition, reduction of the rate of data acquisition from 180 Hz to 60 Hz in the 3-D data sets affected values of kinematic data more than did reduction of 3-D data sets to 2-D (sagittal) coordinates.
Several systematic factors contributed to the differences in the temporospatial and angular variables determined via the 2-D and 3-D systems. The first systematic factor related to the method by which the 2-D system was calibrated. In 2-D systems, kinematic data may be affected by the location of a calibration plane relative to the actual moving plane of the subject and the distance between those planes (perspective error).4 Because the calibration plane was closer to the camera than was the plane of walking on the force plate in our study, the recorded 2-D image may have been exaggerated because the focal length from the lens to the dog was greater than that to the calibration plane. For example, when a hind limb strikes the center of the force plate during walking, the distance between the calibration plane and the plane of walking would be 23.2 cm (half the width of the force plate). When a camera is positioned 3.5 m from the calibration plane, the coordinates of the 2-D markers could be magnified as much as 7%. The second systematic difference concerns the parallax error of the 2-D system, which results when objects move away from the optical axis of the camera.4 In our study, the center of the camera was positioned in the middle of the 2-m calibration area at the height of the stifle joint of the dog. Because we analyzed the 2-D video that contained a complete gait cycle starting from contact with the force plate, the analyzed gait contained images of motion away from the center of the axis, which may have resulted in parallax error. The third systematic difference was attributable to the higher resolution of the 3-D system, compared with that of the 2-D system. This difference of resolution was caused by the technology of the systems and the size of the markers. In our study, the resolution of the 2-D system was 3.8 mm, and that of the 3-D system was 0.1 mm.8 The diameter of markers used with the 2-D system in our study was also approximately 20-fold as large as that of the IREDs used with the 3-D system. Although smaller markers more accurately identify anatomic landmarks, they could not always be identified by the video camera, and therefore, larger markers were required for the 2-D system. The effects of each of the aforementioned factors on the kinematic values obtained by use of each recording system may have been minimal, but the combined effects may have contributed to substantial differences between the 2-D data and the 3-D data.
The effect of eliminating data regarding the mediolateral plane of motion (z coordinate data) was identified by the comparisons between the 3-D and 3-D sagittal data sets as well as the comparison between the 2-D and 3-D data sets. Differences from the comparison of the 3-D 180-Hz and 3-D sagittal 180-Hz data sets were as large as 1.3° for stifle joints and as small as 0.9° for hip joints. The overall effect of reduction of the 3-D data to sagittal (x and y) coordinates was largest in stifle joints, followed by tarsal and hip joints. The differences may have been attributable to the bowed conformation of the hind limbs of dogs in our study, which caused stifle joints to be farthest away from the body, followed by tarsal and hip joints. Thus, motion of hind limbs largely took place in the sagittal versus mediolateral plane for the dogs in our study. However, the effect of mediolateral motion may differ among other breeds of dogs because of differences in the conformations of hind limbs.
In measurements systems with low rates of data acquisition, there is greater chance that an event, such as peak extension or peak flexion of joint, will take place between sequential frames of the recording and will not be detected, compared with results for high rates of data acquisition. The rates of data acquisition of the systems used in our study (60 Hz and 180 Hz) were significantly higher than the minimum rate of data acquisition (17 Hz) calculated by sampling theorem,5 in which 8 Hz was used as the maximum frequency of movement in this study. Thus, the reduction in the rate of data acquisition should not have affected the kinematic data. A significant difference in mean and peak joint angles was detected between the 2-D 60-Hz and 3-D 180-Hz data sets. However, considering that angular excursions of the joints were not different between data sets, these differences may have been attributable to a systematic difference between the 2-D and 3-D data sets rather than a difference in rate of data acquisition. Smoothing of peaks of the data through digital filtering may have also affected the accuracy of kinematic data obtained at a lower rate of acquisition.
The 3-D optoelectric system was used as a gold standard in the study reported here. This system has been used for various analyses of human motion and, with its active markers, is considered the most accurate system.9,10 The active markers are pulsed sequentially so the system can automatically detect each marker by virtue of its pulse timing, and marker tracking is accurately and easily performed.9 Although the accuracy and rates of data acquisition of active marker systems may be superior to those of other kinematic analysis systems, a limitation of active marker systems is the wires, batteries, and pulsing circuitry that must be affixed to a subject, which may interfere with the natural pattern of movement. Therefore, use of 3-D optoelectric systems for kinematic analyses in animals should be limited to evaluations of slow motions in controlled conditions. In the study reported here, limb movement did not appear to be affected by the wires.
The 2-D kinematic analysis system used in the present study is the simplest and most economical 2-D system available and is often used to teach students about kinesiologic principles. Other studies11,12 of animals have been performed with the 2-D configuration and have provided useful information. The 2-D system involves minimal equipment and can be moved easily and installed in locations such as a large hallway or the outdoors. However, a few limitations exist when applying 2-D kinematic analysis to quadruped animals. Because the 2-D system can record only 1 side of motion at a time, motions with an animal moving in opposite directions should be recorded to enable comparison of both sides. These separately recorded motions may not be identical, and larger variability from fatigue that may result from multiple trials may be reflected in the acquired kinematic data. Second, recording of measurement of motion in the frontal plane is difficult because the body moves toward or away from the calibrated plane. Furthermore, accurate recording of limb motions with only 1 camera system is hindered because the forelimbs block the view of the hind limbs.
The limitations of the study reported here include small sample size. With a larger number of dogs, it may have been possible to detect joint angle differences among the 5 data sets. However, more than 100 dogs would have been needed to conduct a study with a power of 80%. Because of the limited availability of the 3-D facility at our institution, we could not evaluate that many dogs. All gait trials were performed within the same day to decrease the variability introduced by change of setting; with more dogs, this would not have been possible.
To our knowledge, the study reported here was the first to compare kinematic data of canine hind limbs during walking obtained via a 2-D video camera system with those obtained via a 3-D optoelectric system. Considering the limitations of 3-D kinematic analysis in veterinary medicine, 2-D kinematic analysis by use of a standard video camera may be a simple, affordable, and accurate alternative to characterize the angular motion of canine limbs in a sagittal plane. The systematic difference between the 2-D and 3-D systems attributable to resolution, marker type, and perspective and parallax errors may be reduced by mathematic corrections. The results of our study suggested that 2-D kinematic analysis may be a valid and repeatable method for assessing the sagittal angular motion of hind limbs of dogs during walking. However, caution should be exercised when linear measurements are evaluated. The 2-D system may be useful for evaluating pathologic limb movements of dogs. Additional evaluation of this system in clinical conditions is indicated.
ABBREVIATIONS
| 2-D | Two-dimensional |
| 3-D | Three-dimensional |
| CV | Coefficient of variation |
| IRED | Infrared light–emitting diode |
Smoothfoam ball, Plasteel Corp, Inkster, Mich.
3M reflective tape, 3M Corp, Saint Paul, Minn.
IRED marker encapsulated small diameter (7 mm), Northern Digital Inc, Waterloo, ON, Canada.
2.4-m twisted pair cable, Northern Digital Inc, Waterloo, ON, Canada.
AMTI OR6-6, AMTI, Watertown, Mass.
Canon GL-2, Canon Inc, Yokohama, Japan.
Omni Light, Lowel-Light Manufacturing, New York, NY.
Optotrak 3020, Northern Digital Inc, Waterloo, ON, Canada.
KAVideo, San Francisco State University, San Francisco, Calif. Available at: www.kavideo.sfsu.edu/Software.htm. Accessed May 7, 2008.
Excel 2003, Microsoft Corp, Redmond, Wash.
Matlab, version 7.0, The Mathworks Inc, Natick, Mass.
SPSS, version 14.0, SPSS Inc, Chicago, Ill.
References
- 1.
Bockstahler BA, Henninger W, Muller M, et al. Influence of borderline hip dysplasia on joint kinematics of clinically sound Belgian Shepherd dogs. Am J Vet Res 2007;68:271–276.
- 2.↑
Hottinger HA, DeCamp CE, Olivier NB, et al. Noninvasive kinematic analysis of the walk in healthy large-breed dogs. Am J Vet Res 1996;57:381–388.
- 3.↑
Kirtley C. Practice—introduction. In: Clinical gait analysis: theory and practice. London: Elsevier Churchill Livingstone, 2006;201–222.
- 4.↑
Kirtley C. Measurement of gait kinematics. In: Clinical gait analysis: theory and practice. London: Elsevier Churchill Livingstone, 2006;39–51.
- 5.↑
Winter DA. Kinematics. Biomechanics and motor control of human movement. 2nd ed. Toronto: Wiley-Interscience, 1990;11–50.
- 6.↑
DeCamp CE, Soutas-Little RW, Hauptman J, et al. Kinematic gait analysis of the trot in healthy greyhounds. Am J Vet Res 1993;54:627–634.
- 7.↑
Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307–310.
- 8.↑
Northern Digital Inc Web site. Optotrak technical specifications. Available at: www.ndigital.com/optotrak-techspecs.php. Accessed Sep 17, 2007.
- 9.↑
Chiari L, Della Croce U, Leardini A, et al. Human movement analysis using stereophotogrammetry. Part 2: instrumental errors. Gait Posture 2005;21:197–211.
- 10.
Ehara Y, Fujimoto H, Miyazaki S, et al. Comparison of the performance of 3D camera systems. Gait Posture 1995;3:166–169.
- 11.
Feeney LC, Lin CF, Marcellin-Little DJ, et al. Validation of twodimensional kinematic analysis of walk and sit-to-stand motions in dogs. Am J Vet Res 2007;68:277–282.
- 12.
Hutchinson JR, Schwerda D, Famini DJ, et al. The locomotor kinematics of Asian and African elephants: changes with speed and size. J Exp Biol 2006;209:3812–3827.
