Comparison of magnetic resonance imaging with standing cervical radiographs for evaluation of vertebral canal stenosis in equine cervical stenotic myelopathy



Reasons for performing study

The sensitivity and specificity of lateral cervical radiographs to evaluate horses suspected of cervical stenotic myelopathy (CSM) are limited by the assessment being restricted to the sagittal plane.


To determine whether magnetic resonance imaging (MRI) allows for a more accurate identification of stenosis than lateral cervical radiographs in horses with CSM.

Study design

Case control study.


Nineteen Thoroughbred horses with CSM (17 males, 2 females, age 6–50 months) were compared to 9 control Thoroughbreds (6 males, 3 females, age 9–67 months). Ante mortem, the subjects had neurological examinations and standing cervical radiographs with sagittal ratios calculated from C3 to C7. Intact cervical column MRI scans and histological examinations of the spinal cord were performed post mortem. Morphometric parameters were measured on the vertebral canal, spinal cord and intervertebral foramen.


Radiographic cervical canal height measurements categorised by standard minimal sagittal diameter intravertebral and intervertebral ratios produced several false positive and false negative determinations of canal stenosis as defined by spinal cord histopathology. Post mortem MRI measurements of canal area and cord canal area ratio more accurately predicted sites of cord compression in CSM cases. No differences in spinal cord measurements were observed when comparing CSM to control horses, but each of the vertebral canal parameters achieved significance at multiple sites.


Vertebral canal area and cord canal area ratio are better parameters to predict the location of cervical canal stenosis compared to only the sagittal plane of canal height. Additional visual planes and measurements obtained by MRI, specifically vertebral canal area and the cord canal area ratio, will provide a more accurate method to identify regions of canal stenosis than lateral cervical radiographs. The development of MRI or computed tomography equipment capable of evaluating the cervical column of mature horses may substantially enhance evaluation of CSM patients.

The Summary is available in Chinese – see Supporting information.


Cervical stenotic myelopathy (CSM, wobbler syndrome) is the result of malformations of the cervical vertebrae, which cause stenosis of the vertebral canal and compression of the spinal cord. Compressive myelopathy has been reported in several species including horses, dogs and man [1-4]. Cervical stenotic myelopathy typically affects young horses, with an increased incidence in certain breeds, including Thoroughbreds, Tennessee Walking Horses, Quarter Horses and Warmbloods [5]. There is a well-documented gender ratio of males over females with reports ranging from 3:1 up to 23:1 [6][7]. Cervical stenotic myelopathy is a multifactorial disease with genetics, high planes of nutrition, alterations in copper and zinc ratios, rapid growth rates and trauma thought to play a potential role in the aetiology and pathogenesis.

Several research groups have made important contributions to the ante mortem diagnostic criteria of equine CSM using standing cervical lateral radiographs and myelography [8-11]. Thresholds have been established using anatomical measurements converted to intravertebral and intervertebral ratios to identify presumptive areas of canal stenosis both at the intervertebral joint and within vertebrae. With myelography, spinal cord compression can be visualised owing to attenuation of the dorsal and/or ventral contrast medium columns and interpreted using published criteria for contrast medium column measurements. Compression from the lateral aspect can also occur due to altered articular process shape, size and spatial positioning. While both methods are important diagnostic aids, the levels of false positives and false negatives remain problematic [12]. A factor contributing to this problem is the limitation of assessment to the sagittal plane with lateral images. Changes in vertebral canal structure that might lead to spinal cord compression are not restricted to reductions in the dimension of dorsal-ventral height. Therefore, a more complete view of the cervical column should improve diagnostic assessment. The ability of magnetic resonance imaging (MRI) to evaluate a structure of interest in multiple planes, as well as provide images of both soft tissues and osseous structures in situ, has the potential to overcome major limitations of lateral radiographs when assessing equine cases for CSM. Given these characteristics, this study was designed to test the hypothesis that MRI provides a more rigorous assessment of the cervical vertebral canal than standing cervical radiographs in the horse. The objectives were: 1) to compare the ability of standing cervical radiographs and MRI to identify vertebral canal stenosis in horses with CSM; and 2) to evaluate the anatomic parameters of the vertebral canal, spinal cord and intervertebral foramen between control and CSM horses on MRI.

Materials and methods


Horses used in the study were identified based on clinical history, neurological assessment, cervical radiographs and post mortem examination (Item S1). Nineteen Thoroughbred horses with CSM (17 males, 2 females, age range 6–50 months with a mean of 18.1 months) were compared to 9 control Thoroughbred horses (6 males, 3 females, age range 9–67 months with a mean of 12.4 months). Case and control horses were identified in collaboration with local veterinary practitioners.

Clinical examination

A neurological examination was performed on each horse (S.M.R.). A grading scale of 0 to 5, where 0 indicated no neurological deficits and 5 indicated recumbency, was used to classify the severity of clinical signs in the forelimbs and hindlimbs [1]. Standing lateral cervical radiographs were performed on all horses. Radiographic images were measured and evaluated by a single blinded assessor (S.M.R.). Standard minimal sagittal diameter (MSD) intravertebral ratios (C2–C5<0.50; C6<0.52; C7<0.56) [9] and intervertebral ratios (C2–C7<0.485) [10] were calculated from C2 to C7 using DICOM Viewer software (eFilm Workstation)a. After ante mortem clinical examination and radiographs, owner-elected euthanasia was performed with an overdose of a barbiturate following AVMA guidelines. Control horses had normal neurological function, but were subjected to euthanasia for other health-related reasons.

Magnetic resonance imaging evaluation

Within 4 h of euthanasia the cervical spinal column was disarticulated in the area of T3 to T5. Loss of cerebral spinal fluid was prevented by placing putty in the vertebral canal at the disarticulation site (Plumber's Putty, model No. 311662)b. Magnetic resonance imaging was performed using a 1.5 Tesla magnet with a flexible surface receiver coil and spinal coil (Siemens Symphony)c. The cervical column was imaged in cranial (C2–5) and caudal (C4–C7) sections. Sagittal, transverse and dorsal plane images using multiple sequences were acquired for each section (Item S2). The cervical column was evaluated in the neutral position. Morphometric measurements were made by 2 independent assessors (K.S.G. and A.P.P.) at the mid-point of intervertebral joints for all parameters using an open source DICOM viewer programd [13]. Measurements were as follows: spinal cord height, spinal cord width, spinal cord circumference, spinal cord area, vertebral canal height, vertebral canal width, vertebral canal circumference, vertebral canal area, left intervertebral foramen height, right intervertebral foramen height, and a ratio of the measured spinal cord area to vertebral canal area (Fig 1).

Figure 1.

Anatomical measurements of the vertebral canal and spinal cord using proton density weighted magnetic resonance images. Transverse plane images demonstrate spinal cord and vertebral canal measurements at the midpoint of the cervical vertebral articulation. a) Height and width measurements of the vertebral canal. b) Vertebral canal area and circumference. c) Spinal cord height and width measurements. d) Spinal cord area and circumference.

Post mortem evaluation

Post mortem examinations were performed by 2 assessors (J.G.J. and N.M.W.). The spinal cord was removed from the vertebral column during disarticulation of intact vertebrae. Intervertebral sites on the spinal cord were identified and marked during disarticulation. The spinal cord was placed in 10% neutral buffered formalin for fixation and processed routinely for histopathological examination. Two to 4 tissue sections were evaluated at each intervertebral site from C1–C2 to C6–C7. Cervical stenotic myelopathy was confirmed by identifying compressive lesions microscopically at specific intervertebral sites based on the characteristic changes of Wallerian degeneration with the presence of gitter cells, dilated myelin sheaths, spheroids (swollen axons) and fibrosis in appropriate dorsal, lateral and ventral funiculi within the white matter [14]. Post mortem and histopathological examinations were also used to rule out other infectious, inflammatory or degenerative neurological diseases of the spinal cord.

Experimental group classification

Ante mortem criteria for CSM horses included: (1) hindlimb ataxia and proprioceptive deficits, and (2) narrowing of the cervical canal on radiographs based on intravertebral (C3-5<0.50; C6<0.52; C7<0.56) or intervertebral (<0.485) minimal sagittal ratios. Histopathological evidence of spinal cord compression on post mortem examination was the definitive parameter used for a CSM diagnosis and for specific site localisation of disease (Table 1). Inclusion criteria for control horses were an absence of neurological deficits on the clinical examination and an absence of compressive lesions on histologic examination of the spinal cord.

Table 1. Study case categorisations based on different radiographic and histological thresholds and assessments
 Standard MSD thresholds for intravertebral sites, and intervertebral sites at 0.485 (radiographs)Intra- and intervertebral thresholds at 0.485 (radiographs)Histopathologic evidence of spinal cord compression (post mortem)
  1. Control (n = 9) and cervical stenotic myelopathy (n = 19) horses were classified at each individual intervertebral site. Some CSM horses were narrowed at multiple sites, accounting for variation in numbers in the cervical stenotic myelopathy category. CSM = cervical stenotic myelopathy; MSD = minimal sagittal diameter.

Data analysis

Vertebral canal measurements at sites of spinal cord compression as identified by histopathology in CSM horses were compared to those of control horses at the same cervical position. Therefore, the total number of CSM samples analysed at each vertebral site was different (Table 1). Data measurements from all 9 controls were used at each vertebral site. Using the Anderson-Darling test, morphometric parameter data were normally distributed. Data for morphometric parameters were analysed using a mixed statistical model that accounted for variables of vertebral site, age, gender, and set using the PROC MIXED procedure in SASe. PROC LOGISTIC was used to determine whether intravertebral and intervertebral radiographic ratios, MRI canal area and cord canal area ratios (CCAR) could localise sites of histopathologic lesions on a vertebral site-by-site comparison. To determine whether canal area and CCAR could predict a lesion regardless of vertebral site, a generalised estimating equation accounting for repeated measures was constructed with the additional variables of age, gender, site and set (Item S3). Evaluation of the 2 MRI image assessors for concordance was done by calculating an intraclass correlation coefficient for each morphometric parameter. Coefficients, using a scale range of 0.00–1.00, were classified as follows: excellent (>0.75), good (0.4–0.75), and poor (<0.4) [15]. The intraclass correlation analysis was performed using MedCalc for Windows, version 12.5f. Significance was defined as P<0.05.


Identification of false positive and false negative cases

Based on the established radiographic MSD ratio and intervertebral thresholds with reference to histopathological assessment of spinal cord compression, a subset of control and CSM horses were categorised as false positives or false negatives (Table 1). Misclassifications based on MSD ratios were observed at each vertebral site and are illustrated in Figure 2 at the level of C4–5. Two control horses were classified as having vertebral canal narrowing at C4–5, whereas 3 of the CSM horses had vertebral canal ratios above thresholds used to indicate a likelihood of stenosis. In contrast, using canal area determined from the MRI images, control and CSM horses segregated more distinctly (Fig 2). Similar partitioning of CSM and control groups based on canal area and CCAR was observed at the other cervical vertebral sites (Items S4 and S5).

Figure 2.

Scatter plot of the minimal sagittal diameter ratios from radiographs and canal area measurements from magnetic resonance imaging of horses with compressive lesions at C4–5. The left ordinate is for the sagittal radiographic ratios with positions of the established thresholds indicated. The right ordinate is for the magnetic resonance imaging canal area data. Two control horses (blue circles) have minimal sagittal diameter ratios indicative of spinal cord compression, which was not present on post mortem histology. Three cervical stenotic myelopathy horses (red circles) fell above the radiographic thresholds, even though spinal cord compression was confirmed by post mortem histology. Canal area values for 2 false positive control and 3 false negative cervical stenotic myelopathy horses (blue and red diagonal lines respectively) cluster within the appropriate groups.

Anatomic changes of the cervical vertebral column on post mortem MRI

Although no significant differences in spinal cord measurements were observed on MRI, each of the vertebral canal parameters achieved significance at multiple sites when comparing CSM to control horses (Table 2, Item S6). The ratio of spinal cord area to vertebral canal area was larger in the CSM group, consistent with a decreased space within the vertebral canal for the spinal cord to occupy. The intraclass correlation coefficient evaluating assessors ranged from 0.70 to 0.89, indicating good to excellent concordance for the morphometric parameter assessment.

Table 2. Comparison of vertebral measurements between control and cervical stenotic myelopathy (CSM) horses on magnetic resonance imaging
SiteNumberCanal height (cm)Canal width (cm)Canal circ. (cm)Canal area (cm2)L IV foramen height (cm)R IV foramen height (cm)Cord area/ canal area ratio
  1. Data are presented as means and standard deviations. Cervical stenotic myelopathy disease status was based on post mortem histopathological evaluation of the spinal cord. *P<0.05; **P<0.01. Site C2–3 had only one cervical stenotic myelopathy horse and could not be statistically evaluated. †indicates significant interaction of age; Circ. = circumference; L IV = left intervertebral; R IV = right intervertebral; CSM = cervical stenotic myelopathy.
C2–3Control (n = 9)2.63 ± 0.522.17 ± 0.208.85 ± 1.295.65 ± 1.310.76 ± 0.150.78 ± 0.170.21 ± 0.07
CSM (n = 1)2.564.4310.107.421.141.260.15
C3–4Control (n = 9)2.18 ± 0.302.44 ± 0.418.00 ± 0.944.62 ± 0.970.88 ± 0.170.88 ± 0.170.28 ± 0.07
CSM (n = 13)1.92 ± 0.28**2.12 ± 0.43**6.94 ± 0.963.49 ± 0.95**0.66 ± 0.23**0.73 ± 0.17**0.33 ± 0.08**
C4–5Control (n = 9)2.14 ± 0.282.59 ± 0.358.10 ± 0.984.70 ± 0.900.91 ± 0.140.95 ± 0.170.25 ± 0.05
CSM (n = 7)1.75 ± 0.28**2.26 ± 0.41**6.75 ± 0.74**3.21 ± 0.66**0.72 ± 0.24**0.74 ± 0.30*0.37 ± 0.08**
C5–6Control (n = 9)2.18 ± 0.262.76 ± 0.278.60 ± 0.565.07 ± 0.581.05 ± 0.110.98 ± 0.170.26 ± 0.03
CSM (n = 11)1.80 ± 0.35**2.51 ± 0.59*7.22 ± 1.38**3.73 ± 1.19**0.75 ± 0.21**0.80 ± 0.23**0.36 ± 0.09**
C6–7Control (n = 9)2.37 ± 0.413.14 ± 0.539.77 ± 01.146.40 ± 1.331.18 ± 0.181.20 ± 0.230.25 ± 0.05
CSM (n = 6)1.85 ± 0.46**2.69 ± 0.55**8.20 ± 1.48**4.38 ± 1.52**0.73 ± 0.30**0.80 ± 0.30**0.38 ± 0.10**

MRI and radiographic measurement assessment for determination of compression

Canal area and CCAR measured on MRI were evaluated for the ability to predict compressive lesions in a site-by-site comparison. All vertebral sites achieved statistical significance with canal area and CCAR (P<0.05) as the predictor of a compressive lesion at a particular vertebral site (Item S7). In contrast, for the radiographic parameters, lesion site significance (P<0.05) was limited to C3–4 and C5–6 for both intravertebral and intervertebral ratios. The MRI canal area parameter (P<0.05) was able to predict compressive lesions across the entire sample set regardless of vertebral site. CCAR was also significant for predicting compressive lesions.


The primary objective of this study was to compare standing cervical radiographs with MRI for the evaluation of cervical vertebral canal stenosis. Magnetic resonance imaging allows examination of both soft tissue and skeletal structures in situ in multiple planes, which addresses the current limitation of only evaluating the sagittal plane when using standing radiography. The ability to assess the cervical column in a transverse plane allows for the application of additional anatomic parameters and a more rigorous evaluation of the variability in cervical vertebral structure, which can cause stenosis in diseased cases. The results indicate that MRI is superior as an imaging tool for the identification of vertebral canal narrowing in horses with CSM compared to standing cervical radiographs. This finding is consistent with the results of a study that examined the use of MRI to assess canine cervical vertebral pathology. Comparison of myelography and MRI techniques in Doberman Pinschers found that MRI had a higher degree of accuracy for identifying the site and severity of spinal cord compression [16]. In the current study, spinal cord compression of CSM horses frequently demonstrated circumferential attenuation of cerebral spinal fluid on transverse images (Fig 3). Significant differences were seen in anatomical measurements of the vertebral canal (Table 2), but not the spinal cord itself. This finding is consistent with a pathogenesis that starts with skeletal malformations resulting in compression of the cervical spinal cord [17].

Figure 3.

Transverse magnetic resonance image showing the absence and presence of spinal cord compression. The image on the left is at the level of C4–5. The spinal cord is round, with circumferential cerebral spinal fluid (green arrows). From the same horse, the image on the right is at the level of C5–6 and demonstrates a narrowed canal area with ventral compression of the spinal cord, attenuated cerebral spinal fluid and encroachment of the articular process joints (yellow arrows). Compression of the spinal cord at C5–6 was confirmed on histological examination.

Interestingly, multiple control horses had at least one area with radiographic measurements suggestive of cervical spinal canal narrowing on standing lateral views, but no neurological deficits on clinical examination or histopathological evidence of spinal cord compression. Figure 2 is a graphic comparison of the corresponding canal areas and standard radiographic ratios for all horses in the study with compressive lesions at C4–5. Using either the standard MSD or 0.485 thresholds for cervical radiography at intravertebral and intervertebral sites, 2 controls fall below these lines and are false positives, given their absence of compressive lesions on histology. In addition, 3 horses with cord compression confirmed by histology have values above these thresholds, making them false negatives on radiographs. Canal area measurements obtained from MR images at the same sites, however, segregate more accurately. As a 2-dimensional transverse measurement, canal area accounts for variability not only in canal height but in other directions as well. Sagittal plane assessments on lateral cervical radiographs are limited to assessment of canal height in the dorsoventral direction only. Variation in vertebral canal shape may provide adequate room for the spinal cord despite a low canal height. This is a possible explanation for the subset of horses with a suggestion of compression on lateral radiographs, yet an absence of clinical or pathological lesions consistent with CSM. Although some ambiguity remains, as represented by an overlap of canal area and CCAR measurements between CSM and control horses at various vertebral sites (Items S4 and S5), the results of this dataset indicate an improvement over lateral cervical radiographs.

Neck position is an additional opportunity for improving the accuracy of diagnostic imaging. Myelograms have been used to demonstrate how cord compression can be accentuated in flexion or extension in some cases [18]. This has led to a description of lesions as dynamic if flexion or extension accentuates the compression, or static if compression is evident regardless of neck position. Dynamic lesions might also explain the apparent paradox of cord canal area ratios being <1 at sites of compression in CSM horses. Although the mean values are significantly higher than the controls, measurements ranging from 0.33 to 0.38 cm2 still suggest adequate space for the spinal cord within the vertebral canal. However, if the compression is position dependent, then the true severity of the stenosis may not be reflected in the CCAR with the neck in a neutral position. There are also varying degrees in the severity of stenosis, and therefore the average CCAR for the sample set does not reflect the gradient of stenosis observed in individual cases.

Intervertebral foramen heights varied at different vertebral sites, but were consistently smaller in CSM horses. This finding supports previous work in which CSM horses were examined by computed tomography (CT) with lateral compression of cervical spinal nerve roots due to malformations in the articular processes [19]. Altered position of the articular processes impacted intervertebral foramen height. Identification of significant differences between CSM and control groups with regard to intervertebral foramen heights is largely dependent on the spatial positioning and size of the articular processes.

As discussed above, the determination of canal area is a more accurate assessment of canal size, but the ability to predict compressive lesions is the important diagnostic issue. In contrast to intervertebral and intravertebral sagittal ratios, canal area and CCAR were able to predict histopathological compressive lesions regardless of vertebral site. Overall, the vertebral canal area was smaller at sites of compression identified by histological examination than in the controls. In the exceptions, where CSM canal areas segregated with the canal area of controls, it is felt that this was largely due to the limitations of only the neutral position being evaluated in this study. Therefore, in the future, canal area and CCAR could be developed as morphometric parameters for ante mortem assessment of CSM in the horse.

Age had an influence on several morphometric parameters in this study. The growth plates of the vertebral endplates in the cervical column do not close until age 4–5 years. Horses included in the sample set were up to age 5.5 years; therefore, there is continued maturation of the vertebrae as the result of normal growth patterns in the horse. These findings reflect the young age of onset often seen in the development of CSM.

The current study was performed with a sample set of 19 CSM horses and 9 controls, yet was still able to detect significant differences between the 2 groups of horses. Examination of more horses in future studies will provide additional power for the establishment of threshold MRI criteria to predict the presence or absence of stenosis, as is currently defined for the assessment of standing cervical radiographs with the intravertebral and intervertebral minimal sagittal diameter ratios. With an increased sample size, a receiver operating characteristic (ROC) curve could be constructed for assessment of diagnostic accuracy and determination of the sensitivity and specificity of the canal area parameter. Given the significance of canal area and CCAR to identify specific sites of compression on post mortem MRI, it is felt these parameters could possibly provide an increased sensitivity and specificity for identifying canal stenosis ante mortem. This could have a strong impact on site selection for cases undergoing review for surgical treatment. Further investigation of the vertebral canal area and CCAR is necessary with regard to the influence of position and the establishment of intervertebral site-specific thresholds. However, as noted by Mitchell et al. [20], breed differences are likely to be important. All horses evaluated in the current sample set, both cases and controls, were Thoroughbreds with an age range of 6 months to 5.5 years. Absence of breed variation probably contributed to the finding that canal area could accurately predict lesion site, in contrast to the earlier study. For this reason, CCAR or another ratio parameter may prove to accommodate breed and age variations better than the absolute measure of vertebral canal area.

Magnetic resonance imaging and CT technology is continuing to advance. Although the size of most mature equine patients currently precludes ante mortem assessment of the complete cervical column by MRI, large bore CT scanners may permit imaging of the entire equine cervical vertebral column. It is feasible that transverse plane vertebral column measurements could be obtained in this fashion for clinical cases. The ability to image the cervical vertebral column in the transverse plane and multiple neck positions will substantially enhance the clinical evaluation of CSM horses.

Authors' declaration of interests

No competing interests have been declared.

Ethical animal research

Procedures were performed with informed consent of the horse owners and under the approval of the University of Kentucky Institutional Animal Care and Use Committee (IACUC).

Source of funding

This research was supported financially by the Grayson-Jockey Club Research Foundation, the Morris Animal Foundation, the Lourie Foundation, and the Maxwell H. Gluck Equine Research Center.


The authors wish to sincerely thank contributing owners and veterinarians for their support of this study. We also gratefully acknowledge the technical assistance of Sara Harpenau, Susan Minnis, Adrial Sitzes, and Emily Zurkuhlen.


J. Janes contributed to the design of the study, acquisition of the data, analysis and interpretation, the drafting and revising of the article. K. Garrett contributed to the design of the study, MR image data acquisition, MR image analysis and revising of the manuscript. K. McQuerry contributed to the design of the study, statistical analysis and revising of the article. A. Pease contributed to the design of the study, MR image analysis and revising of the article. N. Williams contributed to the design of the study, pathology assessment and revising of the article. S. Reed contributed to the design of the study, clinical assessment and revising of the article. J. MacLeod contributed to the design of the study, analysis and interpretation, drafting and revising of the article. All authors gave their final approval of the manuscript.

Manufacturers' addresses

  1. 1Merge eMed, Chicago, Illinois, USA.

  2. 2Oatey, Cleveland, Ohio, USA.

  3. 3Siemens Medical Solutions, Malvern, Pennsylvania, USA.

  4. 4Osirix, Geneva, Switzerland.

  5. 5SAS, Cary, North Carolina, USA.

  6. 6MedCalc Software, Ostend, Belgium.