Dynamic three-dimensional computed tomographic imaging facilitates evaluation of the equine cervical articular process joint in motion

Background: Dynamic computed tomography (CT) imaging has been introduced in human orthopaedics and is continuing to gain popularity. With dynamic CT, video sequences of anatomical structures can be evaluated in motion. Objectives: To investigate the feasibility of dynamic CT for diagnostic imaging of the equine cervical articular process joints (APJs) and to give a detailed description of the APJ movement pattern. Study design: Descriptive cadaver imaging. Methods: Cervical specimens of twelve Warmblood horses were included. A custom-made motorised testing device was used to position and manipulate the neck specimens and


| INTRODUC TI ON
Three-dimensional (3D) reconstruction of computed tomographic (CT) images can be used in presurgical planning and has other applications in veterinary patients. 1Dynamic CT imaging has been introduced in human orthopaedics over the past decade and is continuing to gain popularity, but the technique is not yet utilised in veterinary medicine. 2,3With dynamic CT, anatomical structures can be evaluated in motion as multiple image volumes are recorded over time and subsequently reviewed in form of a video sequence. 40][31][32] Pathology in the equine cervical spine commonly involves a dynamic component.Therefore, assessment of affected segments in motion has the potential to aid the detection of dynamic stenotic conditions.
The aim of this study was to investigate the feasibility of dynamic CT for diagnostic imaging of the equine cervical APJs and to give a detailed description of APJ movement in a cadaver model.We hypothesised that dynamic CT facilitates the comprehensive assessment of the equine cervical APJ components in motion.

| Neck specimens
Cervical specimens (n = 12) were collected from Warmblood horses without evidence of cervical pathology, humanely destroyed for reasons unrelated to this study.Horses were between 3 and 31 years of age (mean = 12 years) and included six geldings and six mares.
This sampling protocol was adopted to comply with the exploratory nature of this study and with animal welfare considerations.
Specimens were excised at the atlantooccipital joint and at the cervicothoracic junction. 24The perivertebral musculature including the intertransversarii cervices and longus colli muscles and the joint capsules were left intact.Specimens were frozen at −20°C immediately after dissection and subsequently thawed over 24 hours at 10°C prior to biomechanical testing.

| Testing device
Based on a previous protocol, a custom-made motorised testing device was used to position and manipulate the neck specimens to perform dynamic computed tomographic (dynamic 2D and 3D CT) imaging (Figure 1). 24The testing device provided consistent and controlled movement of the APJ a detector coverage of 16 cm and a rotation speed of 0.32 seconds provides highquality images of moving objects and gives new insight into the movement pattern of equine cervical APJs.

K E Y W O R D S
articular process joints, cervical spine, dynamic computed tomography, horse, image quality F I G U R E 1 Custom-made motorised device for biomechanical testing of equine neck specimens (C 1 -C 7 ).A specimen is illustrated in neutral position for computed tomographic (CT) imaging (320-detector-row CT scanner, Aquilion One, Canon Medical Systems) of the motion segment C 4/5 (field of view indicated in blue).The body of C 7 is fixed in the device (yellow band).C 1 is secured to a rope, leading through a pulley system and attached to a rope winch in flexion and extension.The vertebral body of C 7 was fixed in the device and C 1 was attached to a rope which led through a pulley and was connected to a rope winch.Additionally, a wooden frame was used to support the weight of the specimen during motion.The motion segment C 4/5 was evaluated in this study.The two motion segments (C 5/6 and C 6/7 ) caudal to the motion segment of interest (C 4/5 ) were not attached to the frame, to allow full ROM of these adjacent segments.The applied force was standardised and limited to 50 Nm and the speed to 0.05 m/seconds. 33

| Computed tomographic imaging
Computed tomographic imaging was performed using an intermittent sequential mode with a 320 detector row CT scanner with a 160 mm wide-area (2D) solid-state detector design that allows image acquisition of a volumetric axial length of 160 mm without moving the CT couch (Aquilion One; Canon Medical Systems).The gantry opening was 78 cm.A field-of-view of 50 cm with 512 × 512 pixels, 0.5 mm slice thickness and a tube rotation time of 0.35 seconds at 100 kVp and 280 mAs (100 effective mAs) were used.Supplementary videos illustrate the CT data (Videos S1-S4): Two dynamic CT scans including full range of extension (Video S1) and flexion (Video S3) and were performed on each neck specimen with focus on the motion segment C 4/5 .

| Computed tomography analysis
The movement of each cervical spine segment (C 4/5 ) was viewed in three planes using multiplanar reconstruction (MPR) mode (dynamic 2D CT) (Figure 2).Sagittal plane images were exported as 2D videos (Videos S1 and S3) as well as dynamic 3D reconstructions (volume rendering) (Videos S2 and S4 and Figure 3).Prior to further evaluation, images were screened for the presence of pathological changes and specimens with evidence of cervical spine pathology were not included in the trial.
For the assessment of the overall image quality and presence of motion artefacts, the dynamic 2D and dynamic 3D videos were divided into four phases of movement.Phases were equally long, with 2-4 seconds per phase depending on the total length of the motion cycle (maximum length 14 seconds).A subjective scale of 1-4 (1 = excellent, no motion artefacts; 2 = good, subtle motion artefacts; 3 = fair, moderate motion artefacts, evaluation possible and 4 = poor, severe motion artefacts, evaluation not possible) was used to grade the overall image quality in each of the four phases (Figure 4).The overall image quality of all four phases was additionally evaluated based on the same score.movements.The intra-observer agreement was assessed using the Wilcoxon Signed Rank Test and the inter-observer agreement was determined using the Friedman test.For all tests, a P-value of <.05 was considered statistically significant.A Bonferroni correction was applied to the significance threshold when relevant.

| Dynamic 2D image quality and artefacts
In flexion, the median image quality of the dynamic 2D imaging was rated with a score of 2 (95% CI: 1-3) in phase II.Subtle motion artefacts were visible but did not impact on the image evaluation.
The median overall image quality of all four phases was 1 in flexion and extension (95% CI: 1-3 both).

| Dynamic 3D image quality and artefacts
For the dynamic 3D videos, all raters gave excellent scores (1) for phases I-IV in flexion and extension.The median overall image quality of all four phases was also rated as 1 in flexion and extension.

| Dynamic 2D vs. dynamic 3D image quality
The image quality of the dynamic 3D videos was significantly better when compared with the dynamic 2D videos in flexion and extension.A significantly lower score was observed in the dynamic 3D videos over the four phases by the three observers for both flexion (95% CI: 1-2 and 1-3, respectively for 3D and 2D videos; P =.007) and extension movement (95% CI: 1-2 and 1-3, respectively for 3D and 2D videos; P =.008).

| Translational displacement of the APJ surfaces
The median length of the cranial and caudal APJ surface was (Figure 5A).In extension, the median displacement of the caudal APJ surface of C 4 relative to the cranial APJ surface of C 5 was 2.78 mm towards caudal for the left side and 1.82 mm for the right side (Figure 5C).There was no statistically significant difference in the displacement of the APJs between the left and right sides of the flexion (95% CI: 12.78-23.00and 12.00-21.87,respectively, for left and right sides; P =.5) and extension (95% CI: 1.21-6.04and 0-8.26, respectively for left and right sides; P =.3)However, significantly more translational displacement was evident in flexion when compared with extension (P =.002 for both the right and the left sides).

| DISCUSS ION
The The volume acquisition speed of the CT is the most important parameter for motion artefact control. 34The volume acquisition speed depends on the gantry rotation speed and the image reconstruction technique. 35An exponential decrease in image quality was seen with lower volume acquisition speeds in previous studies. 2,36Based on these results the highest gantry rotation speed (0.32 seconds per rotation) is recommended when performing dynamic CT examination. 2,36The speed of motion of the biomechanical testing device further influences the development of motion artefacts. 11The speed of the vertebral flexion and extension was pre-determined at 0.05 m/seconds. 24Most likely a slower speed would have resulted in an improved image quality, but also a substantially higher amount of image data as each acquired volume (0.32 seconds) contains 516 images. 37e dynamic 3D videos were free of motion artefacts in phases I-IV in flexion and extension in the current study.This observation can be explained by the calculation of the 3D volume from axial 2D slices, which is accompanied by a loss of resolution.In contrast to another study, there was no evidence for band artefacts in the dataset presented here. 38The wider detector of the CT used (320 detector -row with 160 mm detector width) might mitigate this type of artefact.
Good intra-observer agreement was found for the assessment of dynamic 2D and 3D videos; however, inter-observer agreement was only evident for dynamic 3D but not dynamic 2D video evaluation.
This finding is most likely related to the motion artefacts seen in phases I and II of the 2D videos being very subtle.
The dynamic CT data was used to further investigate the detailed movement pattern of the equine cervical APJs.The translational displacement of the APJ articular surfaces was significantly greater in flexion when compared with extension.In flexion, there was a mean displacement of the APJ surface of over 43% in relation to the mean length of the APJ surface with a displacement of only approximately 7% observed during cervical extension.0][41] Analysis of the dynamic 2D video series of the C 4/5 APJ identified that the cranial margin of C 5 was located in close proximity to the C 4 articular surface in full flexion.Similarly, the caudal margin of C 5 showed what could be interpreted as an area of increased pressure or impingement on the articular surface of C 4 in full extension of the cervical spine, potentially contributing to the development of osteochondrosis or osteoarthritis in this area (Figure 6).In a recent report, osteochondral fragments were identified in 24% of horses with cervical dysfunction. 19The majority of fragments (19/22) were located within the APJ synovial outpouchings, in either a ventral axial or dorsal abaxial location adjacent to the articular margin.
Besides the described musculoskeletal applications, dynamic CT facilitates radiotherapy planning in human patients. 37Targeted radiation can be performed in cases with thoracic neoplasia resulting in a reduced chance of a geographic miss due to movement artefacts. 37,42namic CT angiography is used for visualisation of vascular pathology in the brain and spine and contrast agent enhancement in the heart and pulmonary vessels can be displayed in real time during a cardiac cycle. 37,43,446][47] The ability to display the interaction of articular components during complex movements using dynamic CT additionally facilitates the investigation of potential mechanisms of joint injury.Wide area-detector CT scanners are particularly useful for dynamic studies as they provide high temporal resolution and whole-joint coverage during sequential acquisition. 7Challenges with the widespread use of dynamic CT in routine diagnostic imaging include the cost of the scanner as well as handling and processing of the large volumetric datasets that are generated using this technique. 37 human medicine, the patient is fully conscious and actively performs controlled movements during dynamic CT examination.In veterinary patients, sedation or general anaesthesia is required in order to image controlled movement patterns.In the described investigation, constant and continuous movement of the spinal motion segment C 4/5 was achieved with the aid of a motor-driven external device. in horses when compared with human or small animal patients, the development of similar devices might be useful for the assessment of other regions of the body including the equine distal limb.In humans, however, this CT shows its strength primarily in examinations of anatomical structures in which motion artefacts cannot be avoided (heart and lungs). 37Recently, the possibility of examining limbs of the standing horse (without general anaesthesia) with high-end fan-beam CT systems was introduced. 48Since a distal limb of a standing horse can be examined in 0.32 seconds with this system, significantly less influence of motion artefacts can be expected.
Limitations of the described study are the ex vivo protocol used and the small number of specimens included.Based on previously described protocols, spines were excised at the level of the atlantooccipital and the cervicothoracic junction with only the perivertebral musculature and the joint capsules left intact. 23,49Excision might have influenced the mobility of the spine when compared with the ROM of the equine cervical spine in its entirety.Fixation of the cervical spine specimens in the testing device could have additionally impaired on the motion of the APJs.The motion segments adjacent to C 4/5 were not attached to the frame to minimise the impact of the construct on the cervical ROM.
The described findings should be interpreted carefully with regard to the exploratory nature of the study.The relatively small sample size and the not randomly selected investigated sample may influence results where non-significant differences were found.Additionally, it may not be valid to draw direct conclusions from the data concerning C 4/5 and apply it to other motion segments in the cervical spine as differences in ROM are known to exist particularly towards the cranial cervical spine (C 1 -C 3 ).However, the study is a first step to show the potential of dynamic 3D CT in veterinary medicine.
In conclusion, the results of this study confirm that a CT with a detector coverage of 16 cm and a rotation speed of 0. The study provides a first step in the investigation of the potential of dynamic 3D CT in veterinary medicine, a technique that has only begun to be explored and leaves much room for refinement prior to its introduction in routine practice.
Images were blinded and analysed independently by a boardcertified veterinary radiologist (KV), a board-certified veterinary surgeon (AE) and a third-year equine surgery resident (NS) using open-source imaging software with three-dimensional multiplanar reconstruction (MPR) capability (Horos 64-bit DICOM viewer, Horos Project, 2015) on an image analysis workstation (Apple iMac Pro 2017).The sum of scores attributed by each observer for each of the four phases of movement was used to assess the intra-and inter-observer agreement.The intraobserver agreement was determined by evaluating all images twice within one month (NS) and the evaluation of the inter-observer agreement was based on a comparison of the three observers (KV, AE, NS).For the analysis of the APJ movement pattern, three images were selected for each C 4/5 motion segment, one with the cervical spine in a neutral position, one in maximal flexion and one in maximal F I G U R E 2 Following computer tomographic evaluation, dynamic 2D sagittal plane videos of the C 4/5 motion segment were divided into four phases of movement.Starting at the neutral position (A), 2D captures of initial (B), mid-(C) and full-flexion (D) as well as neutral position (E) initial (F), mid-(G) and full-extension (H) are shown.The four phases of each motion cycle were assessed for overall image quality and presence of motion artefacts extension.The neutral position was defined as the point at which the caudal aspect of the cranial articular process of C 5 and the caudal aspect of the caudal articular process of C 4 were superimposed.In flexion, the cranial translational displacement of the caudal articular process of C 4 relative to the cranial articular process of C 5 was measured.In extension, the caudal translational displacement of the caudal articular process of C 4 compared with the cranial articular process of C 5 was determined (Figure 5).

2. 5 |
Data analysis Data were recorded in Excel (Microsoft Inc) and analysed in SPSS (IBM ® SPSS ® Statistics).Data were visually assessed and found not to be normally distributed and analysed using the Wilcoxon signedrank test for pairwise comparison to compare the difference of translation in flexion vs. extension movements and the difference between the right and left translations both in flexion and extension F I G U R E 3 Sagittal plane computer tomographic images were exported as dynamic 3D reconstructions (volume rendering).The dynamic 3D videos of the C 4/5 motion segment were split into four phases of flexion (A-D) and extension (E-H) movement and graded to describe the presence of motion artefacts and the image quality Four phases of a movement cycle (flexion and extension) were assessed based on video sequences (2D and 3D) of the dynamic computer tomographic evaluation of the C 4/5 motion segment of 12 equine cervical spine specimens.The presence of motion artefacts and the overall image quality were subjectively graded as 1 = excellent, no motion artefacts (A); 2 = good, subtle motion artefacts (B); 3 = fair, moderate motion artefacts, evaluation possible (C) and 4 = poor, severe motion artefacts, evaluation not possible (D) by threeblinded observers

43 .
57 mm.In flexion, the median displacement of the caudal APJ surface of C 4 relative to the cranial APJ surface of C 5 was 18.79 mm towards cranial for the left side and 17.22 mm for the right side F I G U R E 5 (A) Two-dimensional sagittal plane computer tomographic (CT) image of C 4 / 5 in maximal flexion.The translational displacement of the caudal articular process joint (APJ) surface of C 4 against the cranial APJ surface of C 5 towards cranial was measured as indicated.yellow line = length of the caudal APJ surface of C 4 ; blue right angle = translational displacement in flexion (a).(B) Twodimensional sagittal plane CT image of C 4/5 in neutral position with the caudal aspect of the APJ surface of C 4 and C 5 superimposed.(C) Two-dimensional sagittal plane CT image of C 4/5 in maximal extension.The translational displacement of the caudal APJ surface of C 4 against the cranial APJ surface of C 5 towards caudal was determined as shown.yellow line = length of the cranial APJ surface of C 5 ; blue right angle = translational displacement in extension (b) results of this study confirm the feasibility of dynamic 2D and 3D CT for the assessment of the equine cervical APJs based on a cadaver model.Additionally, the investigation gives new insight in the dynamics of the APJ articular components during motion.Blinded evaluation of 2D and 3D video sequences derived from sagittal plane CT data focused on the motion segment C 4/5 identified an excellent overall image quality for both, flexion and extension movement.During dynamic 2D CT imaging, there was of slight motion artefacts adjacent to the APJs in movement phases I and II in flexion and extension.Most artefacts were observed in the vicinity of the APJs as the soft tissues showed a higher velocity of movement than the APJs themselves.Whilst motion artefacts can impair on image analysis and post-processing, it has been demonstrated that the assessment of axial CT sequences is not qualitatively affected by object motion at 0.05 m/seconds. 2In addition, each dynamic CT study consists of multiple individual axial sequences, at least one of which was always free of motion artefacts and could be used diagnostically for multiplanar reconstruction without restriction.
32 seconds makes it possible to obtain high-quality images of equine cervical articular process joint C 4/5 in motion.As the acquired dynamic data sets consist of multiple individual volume sets, slight motion artefacts in individual video sequences can be replaced by a different volume for the evaluation of stationary (2D) images.This work further confirms that dynamic CT imaging has the potential to provide new insight into the movement pattern of articular motion segments, which may be of value for the examination of APJ and other joint conditions.