Coxofemoral joint kinematics using video fluoroscopic images of treadmill-walking cats: development of a technique to assess osteoarthritis-associated disability

Abstract

The objectives of this pilot study were to develop a video fluoroscopy kinematics method for the assessment of the coxofemoral joint in cats with and without osteoarthritis (OA)-associated disability. Two non-OA cats and four cats affected by coxofemoral OA were evaluated by video fluoroscopy. Video fluoroscopic images of the coxofemoral joints were captured at 120 frames/s using a customized C-arm X-ray system while cats walked freely on a treadmill at 0.4 m/s. The angle patterns over time of the coxofemoral joints were extracted using a graphic user interface following four steps: (i) correction for image distortion; (ii) image denoising and contrast enhancement; (iii) frame-to-frame anatomical marker identification; and (iv) statistical gait analysis. Reliability analysis was performed. The cats with OA presented greater intra-subject stride and gait cycle variability. Three cats with OA presented a left–right asymmetry in the range of movement of the coxofemoral joint angle in the sagittal plane (two with no overlap of the 95% confidence interval, and one with only a slight overlap) consistent with their painful OA joint, and a longer gait cycle duration. Reliability analysis revealed an absolute variation in the coxofemoral joint angle of 2º–6º, indicating that the two-dimensional video fluoroscopy technique provided reliable data. Improvement of this method is recommended: variability would likely be reduced if a larger field of view could be recorded, allowing the identification and tracking of each femoral axis, rather than the trochanter landmarks. The range of movement of the coxofemoral joint has the potential to be an objective marker of OA-associated disability.

Introduction

Osteoarthritis (OA) is the most frequent musculoskeletal disease in companion animals, and leads to joint failure and chronic pain.^1–4 In cats, OA prevalence increases with age,^5–9 and it affects quality of life. Several studies suggest that OA is associated with pain, expressed as a decrease in daily activity, reluctance to jump or climb stairs, and other altered behaviors.^9–12 Currently, there is no validated technique to evaluate chronic pain in cats. This can be explained by a combination of the well-known propensity of cats to be poorly expressive of pain, and the complexity of chronic pain assessment related to the development of concurrent physical and cognitive–affective disability.^13,14

The use of gait measurements may provide objective chronic pain outcomes. Telemetered motor activity assessment, which provides an objective quantification of activity limitations, seems to be one of the most promising markers of OA chronic pain-related disability in companion animals.^15–17 Objective quantification of limb impairment using kinetics to evaluate gait profile (mostly peak vertical ground reaction force and vertical impulse) has been effectively applied to evaluate the outcome of OA therapy in dogs.^17–20 In cats, this technique appears to have great potential to detect OA chronic pain-related disability, but needs further validation.^21–26

Kinematics, the study of the motion of a body, can be used to determine information complementary to kinetics for the study of gait, such as velocity, acceleration and joint angles. Video fluoroscopy techniques were developed using a high-speed digital camera and continuous X-rays emission to film dogs and rats walking on a treadmill.^27,28 These techniques provide direct, precise and fluid visualization of bone structures during movement. Extracting the gait pattern data from these sequences involves image pre-processing (image distortion correction, denoising and contrast enhancement), tracking of morphological markers in video frames and statistical gait analysis.

The objectives of this study were to evaluate the feasibility of video-fluoroscopic gait assessment of the coxo-femoral joint in non-OA and OA cats, and to describe the changes associated with OA chronic pain-related disability.

Materials and methods

Selection and characteristics of cats

The Institutional Animal Care and Use Committee approved the study protocol (# Rech-1482), and the Canadian Council on Animal Care guidelines were followed with respect to the care and handling of the cats. Six cats were selected from our cat colony based on their ability to walk consistently and comfortably on a motor-driven treadmill. The cats were housed in a designated room as previously described. ²¹

The cats presented a normal neurological evaluation, complete blood count, blood chemistry and urine analysis. They also had normal computed radiographs (CRs) of mediolateral and caudocranial projections of the stifle, lumbosacral and sacroiliac joints, as well as mediolateral projections of the shoulders and elbows. All radiographs were performed under sedation with medetomidine (0.02 mg/kg; Domitor 1 mg/ml, Pfizer Canada Animal Health) and morphine (0.1–0.2 mg/kg; Morphine Sulfate Injection 10 mg/ml, Sandoz) administered intramuscularly. The presence of OA was first evaluated using structural assessment of CR and magnetic resonance (MR) images. The cats had not been administered anti-inflammatory or structure-modifying (eg glucosamine, chondroitin or bisphosphonate) medication for at least 3 months prior to the start of the study.

The subjective presence of joint pain during the orthopedic examination just prior to the video fluoroscopic examination completed characterization of the OA. A joint was considered painful if the cat presented consistently (similar response during two consecutive manipulations of the joint) at least one of the following behaviors during the manipulation of the joint: withdrawal, attempting to escape, hissing or biting, body tensing or immobilization, or vocalization.^15,29,30 Non-OA cats (n = 2, aged 1.5 and 2 years old, respectively) presented absence of any sign of OA in both coxofemoral joints. Cats with OA (n = 4; mean age 9 years, range 5–11 years) presented OA of at least one coxofemoral joint (Table 1).

Table 1.

Structural imaging scores, subjective pain and kinematic variables of the left (L) and right (R) coxofemoral joints extracted from the video frames

		Non-OA cats		OA cats
		H1	H2	OA1	OA2	OA3	OA4
Structural scores (radiographic/MRI)	L	0/0	0/0	0/1	2/3	2/2	2/2
	R	0/0	0/0	0/2	2/5	2/3	0/1
Presence of joint pain	L	No	No	Yes	Yes	No	Yes
	R	No	No	No	Yes	Yes	No
Mean coxofemoral joint angle (º) range of movement (95% CI)	L	87.0	63.1	62.45	69.7	80.6	67.4
		(82.6–91.4)	(59.0–67.2)	(57.1–67.8)	(64.5–74.9)	(77.9–83.3)	(64.0–70.8)
	R	79.8	65.2	98.6	67.5	61.1	72.3
		(74.1–85.5)	(61.6–68.9)	(92.7–104.5)	(60.8–74.1)	(58.7–63.5)	(69.4–75.2)
Mean duration of gait cycle, s (95% CI)	L	0.88	0.83	1.17	0.93	1.06	1.0
		(0.86–0.90)	(0.81–0.85)	(1.10–1.24)	(0.87–0.99)	(1.00–1.11)	(0.94–1.06)
	R	0.88	0.825	1.18	0.95	1.04	0.975
		(0.86–0.89)	(0.80–0.85)	(1.12–1.23)	(0.891.01)	(1.00–1.07)	(0.94–1.01)
Mean percentage of gait cycle reached at the maximum, % (95% CI)	L	37.9	36.5	31.4	39.9	37.6	34.8
		(36.4–39.3)	(33.8–39.2)	(28.6–34.2)	(37.4–42.3)	(35.4–39.7)	(32.1–37.5)
	R	39.1	37.6	34.2	41.6	48.8	42.1
		(36.5–41.6)	(34.4–40.8)	(32.2–36.2)	(38.7–42.3)	(45.5–52.0)	(39.5–44.6)

OA = osteoarthritis; MRI = magnetic resonance imaging; CI = confidence interval

The detailed methodology for imaging modalities used to evaluate OA has been reported previously. ²¹ Briefly, the coxofemoral joints were imaged using mediolateral and ventrodorsal CR projections (CR-DX system; Agfa), and using a 1.5-Tesla MR imager (HDx; GE) and an eight-channel human-knee coil under general anesthesia. The CR and MR images were evaluated under blinded conditions to assess structural changes (osteophytes, enthesophytes, subchondral bone sclerosis [CR and MR images], joint effusion, bone marrow edema-like lesions and femoral head cartilage thinning [MR images]).

Video fluoroscopic acquisition of the coxofemoral joints

The cats were progressively trained to walk on a motor-driven treadmill at a speed of 0.4 m/s inside an open-top, transparent Plexiglas box over a 6 week period. After this training, the video-fluoroscopic sequences of the coxofemoral joints during locomotion were acquired using a C-arm X-ray system (Coroskop, Siemens, Dorval, QC, Canada) equipped with an image intensifier with an operating field of view of 27 cm (diameter) and a motor-driven treadmill mounted on a triaxial free-moving table. Most of the commercial fluoroscopic units sample at a maximum frequency of 25 frames/s, which is insufficient to provide low noise and sharp images of a bone during movement. This limitation is owing, in part, to pulse X-ray fluoroscopy, which has been developed to decrease the dose of radiation associated with fluoroscopic examinations. ³¹ To record adequately joint motion, the commercial C-arm fluoroscopic unit was customized to obtain an acquisition frequency of 120 frames/s. The video fluoroscopic sequences were recorded using a high-speed scan camera (DS-41-300K0262; Dalsa) equipped with a C-mount zoom lens (Fujinon-tv H6X12.R, 1:1.2/12.5-75; Fujifilm North America) fixed on the image intensifier. High-speed, real-time imaging system software (VisionNow; Boulder Imaging) supported the camera controls, image captures and streaming to disks. Video fluoroscopic images were captured while a cat walked freely (but was restrained in the open-top Plexiglas box) at a speed of 0.4 m/s, imposed by the treadmill (see Video S1 in Supplementary material). The cat’s coxofemoral joints were kept in the center of the image intensifier by manually moving the table. Raw digital video fluoroscopic images (512 × 512 pixels with 256 gray levels) were acquired at 120 frames/s using a shutter speed of 2 ms and an X-ray output setting of 90 kV/13 mAs. For each cat, two sessions of a maximum of 15 mins duration each were used to obtain a maximum of step cycles, while avoiding exercise-related stress and fatigue.

A single frame of a video of a known 27 × 27 cm calibration grid positioned in the field at the same location as the cat was acquired at the beginning of the session to allow further image correction. Because the distance from X-ray source to image intensifier, field of view, X-ray output settings and camera zooming were not modified between cats, this calibration frame was valid for the whole experiment. After the acquisition, each raw video fluoroscopic sequence was visualized using the above mentioned real-time imaging software and segmented into shorter sequences of two or more successive steps, the whole coxofemoral joint being present in all the images of the sequence. Then, those sequences (the addition of different sequences of two or more successive steps providing a total of 18–24 steps for each cat) were converted to an audio video interleave (AVI) uncompressed file for analysis (see Figure 1a and Video S2 in Supplementary material).

Figure 1.

Pre-processing steps of the video fluoroscopic imaging. (a) Raw frame; (b) frame after correction for distortion, image denoising and contrast enhancement; (c) frame with anatomic markers placed on the lesser trochanter (square markers), the femoral head (circle markers), the cranial dorsal iliac spine (star markers) and the ischiatic tuberosity (triangle markers) of the left (yellow markers) and the right (red markers) hindlimbs; (d) illustration of the left coxofemoral joint angle (angle 1) and coxofemoral bone angle (angle 2)

Pre-processing and kinematic data of video fluoroscopic imaging

All operations on video fluoroscopic sequences were performed using an in-house graphic user interface (GUI). The GUI allowed four operations: (i) correction for image distortion, (ii) image denoising and contrast enhancement, (iii) frame-to-frame morphological marker identification, and (iv) statistical gait analysis.

First, geometric image distortions caused by the X-ray imaging chain (generally variation in magnification of 5–10%) were measured using the rectilinear calibration grid, and corrected using a standard computer procedure of inverse transformation and image interpolation methods for two-dimensional (2D) images. ²⁸ Second, to obtain the best bone VS soft tissues image contrast, each video frame was denoised and contrast-enhanced using wavelet-based denoising methods (see Figure 1b and Video S3 in Supplementary material). ³² Third, for each hindlimb and each frame, a marker was placed on the lesser trochanter, the femoral head, the cranial dorsal iliac spine and the ischiatic tuberosity using a semi-automated method (see Figure 1c and Video S4 in Supplementary material). Briefly, using a computer mouse, for each limb separately, the user identified the markers on the frames with minimal superposition, and the GUI estimated their positions in the other frames. Most often, this approach allows the user to locate the markers in about one image out of six, but in one image out of three during the high acceleration phases of the movement. Then, the 2D anatomical markers trajectories across each video sequence were processed and analyzed to calculate coxofemoral joint angle changes over time.

The coxofemoral joint angle was defined using the lesser trochanter, femoral head and caudal dorsal iliac spine markers (angle 1 in Figure 1d). The coxofemoral bone angle was defined using the cranial dorsal iliac spine, femoral head and ischiatic tuberosity markers (angle 2 in in Figure 1d). Analysis of the joint angle pattern over time used 2D trajectories of the markers by time, from which the projected 2D angles in the sagittal plane were computed, allowing the gait patterns to be estimated. ²⁸ The coxofemoral bone angle should be constant over time in the sagittal plane as the centre of the coxofemoral joint is assumed to be constant throughout the gait cycle. Thus, this latter angle provided a reliability estimator (see data analysis). For each hindlimb, the onset of the gait cycle (approximate to paw contact) was set when the coxofemoral joint angle reached its minimum.

Data analysis

For each hindlimb 18–24 steps were analyzed to compute angles over the percentage of gait cycle for both angles previously defined. Then, angles were expressed using mean values. Variables extracted for each cycle consisted of the gait cycle duration (time to complete a whole gait cycle), the range of movement of the coxofemoral joint angle (difference between the maximum and the minimum reached during a step), and the percentage of gait cycle value where the maximum angle was reached. The kinematic variables were expressed as the mean and 95% confidence interval (CI). A cat was classified as asymmetric for a variable when the 95% CIs were distinct between both hindlimbs. Owing to the nature of the study (preliminary study to generate hypotheses), no inferential statistics were performed.

The absolute difference between the left and right coxofemoral joint angles in the sagittal plane was calculated for each cat using the mean value of the left and right coxofemoral joint angle over the gait cycle. For each cat, a step (right or left hindlimb) was randomly selected to provide data for a test–retest reliability analysis. For this analysis, coxofemoral joint angles were computed six times from the markers identification step (three times the same day and three times a week apart). The reliability of the coxofemoral joint angle was expressed for each cat across a gait cycle as SD (absolute variation) and coefficient of variation (relative variation). The coxofemoral bone angle provided a reliability estimator expressed as variation over time using SD of the measure along the gait cycle (absolute variation), and the associated coefficient of variation (relative variation).

Results

The mean coxofemoral joint angle in the sagittal plane across time expressed as percentage of gait cycle presented as a bell-shaped curve for all cats, and a maximum reached at 31–49% of the gait cycle. The global shape of the left (Figure 2a) compared with the right (Figure 2b) coxofemoral joint angle appeared less variable when both hips were functionally similar (the two healthy non-OA cats, H1 and H2, and cat OA2) compared with when hips were functionally different (cats OA1, OA3 and OA4 with unilateral painful OA). This observation was confirmed by visualizing the absolute difference in the coxofemoral joint angle, which showed that cats OA1, OA3 and OA4 clearly presented more asymmetry than cats H1, H2 and OA2 (Figure 3).

Figure 2.

Left (a) and right (b) coxofemoral joint angle in the sagittal plane by percentage of gait cycle. For each cat, the data depicted are mean values calculated from the 18–24 steps analyzed. Black lines depict the osteoarthritic cats OA1 (solid line), OA2 (dotted line), OA3 (dashed line) and OA4 (dash-dot-dot line); gray lines depict the non-osteoarthritic cats H1 (solid line) and H2 (dotted line)

Figure 3.

Absolute difference between the left and right coxofemoral joint angles in the sagittal plane by percentage of gait cycle. For each cat, data depicted are mean values calculated from the 18–24 steps analyzed. Black lines depict the osteoarthritic cats OA1 (solid line), OA2 (dotted line), OA3 (dashed line) and OA4 (dash-dot-dot line); gray lines depict the non-osteoarthritic cats H1 (solid line) and H2 (dotted line)

Cats OA1 and OA3 presented a very marked coxofemoral joint flexion of the right and left side, respectively (<30º), at the beginning of stance and end of swing (Figure 2). Interestingly, these two cats presented significant asymmetry of the mean range of movement of the coxo-femoral joint angle in the sagittal plane, such that the lower range of movement was matched by the painful joint. The left hindlimb of cat OA4 was painful, but there was only a tendency for asymmetry of the joint range of movement (there was only a slight overlap of the two 95% CIs; Table 1). For all cats, the mean gait cycle duration was similar for the right and left coxofemoral joints, but the variability of this parameter was higher in the four OA cats compared with the two non-OA cats (Table 1). Moreover, cats OA1, OA3 and OA4 presented a longer gait cycle duration than the non-OA cats (distinct 95% CI). Furthermore, the coxofemoral joint angle reached its maximum in the sagittal plane at the same time (of the gait cycle) for the right and left coxofemoral joints, except for cats OA3 and OA4 (Table 1). Only cat OA1 reached its maximum at a lower percentage of the gait cycle.

In general, the test–retest analysis showed an absolute variation of the coxofemoral joint angle by 2–6º (Figure 4a), which represented 2–10% of relative variation (Figure 4b). Interestingly, there was a higher variation for all cats at around 10% and from 70% to the end of the gait cycle (see Figure 4b).

Figure 4.

Test–retest observed absolute (expressed as SD in [a]) and relative (expressed as coefficient of variation in [b]) variation of the coxofemoral joint angle in the sagittal plane by percentage of the gait cycle. For each cat, data from a randomly selected step were used to compute coxofemoral joint angles six times. Black lines depict the osteoarthritic cats OA1 (solid line), OA2 (dotted line), OA3 (dashed line) and OA4 (dash-dot-dot line); gray lines depict the non-osteoarthritic cats H1 (solid line) and H2 (dotted line)

The coxofemoral bone angle presented a 5–15º variation along the gait cycle, and a general U-shape-like curve with minima at about 50–55% of the gait cycle (Figure 5a). This U-shape was particularly pronounced for the right coxofemoral bone angle of cat OA1 (Figure 5b) and the left coxofemoral bone angle of cat OA4 (Figure 5a).

Figure 5.

(a) Left and (b) right coxofemoral bone angles in the sagittal plane by percentage of gait cycle. For each cat, data depicted are mean values calculated from the 18–24 steps analyzed. Black lines depict the osteoarthritic cats OA1 (solid line), OA2 (dotted line), OA3 (dashed line) and OA4 (dash-dot-dot line); gray lines depict the non-osteoarthritic cats H1 (solid line) and H2 (dotted line)

Discussion

This study is the first report of an evaluation of the 2D video fluoroscopic coxofemoral joint motion in cats. A video fluoroscopic method was implemented to obtain accurate and non-invasive in vivo joint kinematics. Biplane fluoroscopy, which has provided dynamic measurements of the knee joints in dogs and several joints in humans,^33–37 is currently the state-of-the-art in X-ray fluoroscopy. ³⁸ While biplane methods have been shown to be more accurate, single-plane methods are accurate to 2 mm of translation and 2º of angulation.^39,40 We considered that using single plane fluoroscopy was easier to implement than biplane fluoroscopy, and that it presented appropriate accuracy for our clinical research objective. Video-based system tracking markers attached to the surface of the skin have been used to detect disabilities associated with musculoskeletal injury in cats.^25,41 However, we felt that using video fluoroscopy was a better choice in order to avoid artefacts associated with video-based system tracking markers. Indeed, in humans, large skin motion artefacts were reported, resulting in poor approximations of the kinematics of the underlying bones.^38,42

The coxofemoral joint angle trajectories obtained in this study provided high-quality data that could be used as quantitative markers of motor impairment. This result is owing to the combination of a careful progressive training of the cats to avoid fear of the treadmill and obtain a normal constant walking pace, and the use of the GUI. Twelve cats were trained, but only six completed training; of those excluded, three refused to walk and three walked with a non-constant speed (cyclic run and stop pattern). The GUI was fundamental for performing image pre-processing (correction for image distortion, image denoising and contrast enhancement), and also decreased the number of frames to process. Because of the trajectory interpolation, only 20% of the marker positions needed to be identified by the operator. Without the GUI, the marker positioning on each frame of one video sequence would require between 3 and 4 h to complete (instead of 30 mins). Furthermore, the GUI decreased the difficulty of marker placement by resolving many problems related to superposition.

In this study, an asymmetry between the left and right coxofemoral joint angles was present in three of the OA cats (for cats OA1 and OA3, there was no overlap of the 95% CI between sides, and only a slight overlap for cat OA4). This lower range of movement in the sagittal plane was present on the side detected to be painful during the orthopedic examination. This result is in accordance with a recent study of experimental OA in rats using 2D fluoroscopy, which demonstrated a decreased range of motion in the sagittal plane of painful OA knee joints compared with the contralateral control side. ⁴³ A recent study suggests a tendency of certain joints afflicted by OA to present with a decreased range of motion, but not the coxofemoral joint. ⁴⁴ However, in this study, painful and non-painful joints were evaluated together, which could have decreased the ability to detect differences.

The observed decreased angle of the painful joints in the OA cat group may be related to a protection of this joint during dynamic weight bearing. This could be associated with clinical gait abnormalities that can be detected by owners or veterinarians, as previously reported.^{9,10,12,45,46} This gait modification is also in accordance with the decrease in peak vertical ground reaction force associated with painful OA in cats. ²² In addition, there was more variability in data from the OA cats compared with the non-OA cats, particularly with respect to the shape of the coxofemoral joint angle curve. There was also an increase in gait cycle duration for three of the OA cats. Taken together, these observations may relate to the greater difficulty of OA cats in walking consistently owing to muscle fatigue or pain. This is in agreement with frequent reports of activity and mobility changes in cats afflicted with OA; for example, the presence of lameness, stiffness, decreased ability to jump, reduced motor activity, difficulty climbing or descending stairs, and decreased interest in playing with toys, owners or other animals.^{9,10,12,22,47} The only bilaterally affected cat in this pilot study, cat OA2, did not present a kinematic pattern similar to the three other OA cats (unilateral painful OA). This could be related to a bilateral decrease in the coxofemoral joint angle range of movement. However, a larger sample of non-OA cats would be required to determine the normal joint range of movement in cats.

This study showed good test–retest reliability. The higher variability observed at about 10% and from 70% to the end of the gait cycle may be related to the greater angular velocity during these periods, which would lead to inaccuracies in locating the markers, even with the interpolation method used. This could be associated with a much higher variability of the coxofemoral bone angle. This variation is almost twofold higher than the test–retest observed absolute variation, and is likely due to a rotation of the whole pelvis. Hence, better precision and accuracy of the coxofemoral joint movement could be obtained using a three-dimensional motion analysis methodology as reported in the literature. ³⁸ One aspect of the method described herein that likely contributed to the variability is that the points used to define the position of the femur were close together and were not always distinct on the images. A better precision of the coxofemoral joint angle could have been obtained using a femoral marker farther from the coxofemoral articulation than the lesser trochanter or using a straight line along the femur from the femoral head. However, the lesser trochanter was defined as the best anatomical landmark to use on the femur during the preliminary phase of the study because it was easier to detect and almost always present in the images, in contrast to the distal part of the femur. Unfortunately, the positioning of a straight line was not possible with the design of the GUI.

Another limitation of this study was the impossibility of monitoring either foot location, or the maximum extension of the stifle and tarsus because of the limited field of view. This led to an incapacity to determine stance and swing times, as well as stance and swing velocity and acceleration. Using a C-arm equipped with an image intensifier with an operating field of view with a diameter of 40 cm would overcome this limitation.

Conclusions

This study provides the first evaluation of 2D video fluoroscopic coxofemoral joint motion in non-OA and OA cats. The range of movement of the coxofemoral joint could be an objective marker of OA-associated pain in cats. This should be confirmed in a larger study comparing non-OA and OA (unilateral and bilateral, single or multiple joints affected) cats walking on a treadmill under video fluoroscopy surveillance before and after an analgesic treatment. This would allow determination of whether the asymmetry of the coxofemoral joint range of movement observed in this study is related more to pain or to biomechanical factors (ie, periarticular fibrosis and osteophytes). A larger study should also verify the relationship between age and change in kinematics, as aging could be a confounding factor in OA-associated disability. Because the majority of the coxofemoral joint movement occurs in the sagittal plane, a 2D method with the proposed GUI should be sufficient. However, owing to the limitations of the method in this study, we recommend improvement of the method with the use of a larger field of view and a line defining the long axis of the femur from the femoral head instead of the lesser trochanter marker.