We have the following interests: Clare Rusbridge is employed by Fitzpatrick Referrals and the University of Surrey and Susan P Knowler works voluntarily and is affiliated with Fitzpatrick Referrals. Angus K McFadyen is an independent statistical consultant operating professionally as a sole trader under the company name of akm-stats. There are no patents, products in development or marketed products to declare. This does not alter our adherence to all the PLOS ONE policies on sharing data and materials, as detailed online in the guide for authors.
Conceived and designed the experiments: SPK CR. Performed the experiments: SPK CR. Analyzed the data: SPK AKM. Contributed reagents/materials/analysis tools: ZK CF MK SRP AKM SPK CR. Wrote the paper: SPK CR AKM. Contributed to the original concept of a QTL analysis: ZK.
This study aimed to develop a system of quantitative analysis of canine Chiari-like malformation and syringomyelia on variable quality MRI. We made a series of measurements from magnetic resonance DICOM images from Griffon Bruxellois dogs with and without Chiari-like malformation and syringomyelia and identified several significant variables. We found that in the Griffon Bruxellois dog, Chiari-like malformation is characterized by an apparent shortening of the entire cranial base and possibly by increased proximity of the atlas to the occiput. As a compensatory change, there appears to be an increased height of the rostral cranial cavity with lengthening of the dorsal cranial vault and considerable reorganization of the brain parenchyma including ventral deviation of the olfactory bulbs and rostral invagination of the cerebellum under the occipital lobes.
Canine Chiari-like malformation (CM) is a complex abnormality characterized by overcrowding of the craniocervical junction and by disparity in size between the brain parenchyma (too big) and the skull (too small)
CM is assumed to be a developmental disorder. The rostral portion of the cranial base is of neural crest origin whereas the caudal portion is of mesodermal origin
CM is a complex disorder and although there is less phenotypic variation than with humans, there can be differences between breeds and individuals within the same breed. In particular the conformation of the craniocervical junction varies and in some individuals the size of cerebellar herniation may be minimal
These variations in morphology, and the undetermined significance of their contribution to CM, compound the difficulty in diagnosing the condition and accurate assessment of the risk of a dog developing syringomyelia and/or passing on this risk to offspring. It also creates difficulties in accurate phenotyping for genetic studies.
The Griffon Bruxellois (known as the Brussels Griffon in the USA) has a high prevalence of the disorder with a conservative estimate of 65% having CM and 42–52% having syringomyelia
The diagnosis of CM and syringomyelia is dependent on magnetic resonance imaging (MRI) of the brain and cervical spinal cord. In ideal circumstances, an experienced operator would image all the subjects in a high field MRI unit obtaining excellent quality images enabling a quantitative assessment of the morphology. In reality the cohort for this group’s genetic studies is sourced from a global population of dogs undergoing imaging for diagnostic and screening prior to breeding purposes. Although the dogs are imaged with the same basic protocol, the tesla strength of the MRI unit and the skill of the operators have varied considerably. Another difficulty is that for economic reasons, assessment of breeding dogs is made from only 3 MRI sequences of a limited anatomical area that does not include the entire cranial fossa
This study did not involve live animals but the analysis of MRI DICOM obtained for diagnostic purposes or for determining CM and syringomyelia status. Full consent was obtained from all owners and actual dogs remained anonymous.
The investigation reviewed 155 Digital Imaging and Communications in Medicine (DICOM) MRIs of the brain and cervical region of Griffon Bruxellois dogs. In addition, images from 6 other dogs were measured to provide a comparison for the study. This comprised of 3 mesaticephalic dogs (Beagle and 2 Australian terriers) and 3 Affenpinschers, a brachycephalic toy dog breed that is genetically close to the Griffon Bruxellois. The images had been obtained for diagnostic reasons (for example in the investigation for neurological signs and pain), for screening prior to breeding
The upper image has been “windowed” to create improved contrast and highlight bony landmarks with a magnification of the area of the optic nerve and canal (inset). The lower image demonstrates the same without the windowing effect. All line measurements start from one of 9 points: (A) dorsum of spheno-occipital synchondrosis. (B) basion of basioccipital bone. (C) rostral edge of the dorsal lamina of the atlas. (D) junction between the supraoccipital bone and the occipital crest. (E) most dorsal point of intersection of the cerebellum with the occipital lobe circle. (F) center of occipital lobe circle placed on the cranial baseline (HAI) and extending to encompass the occipital lobes. The centre of the circle is F. (G) point at which the optic nerve deviates into the canal (inset). (H) most caudal point of the olfactory bulb. (I) intersection point with the extended HA baseline. 5 angles measured are (1) ABC, (2) CAF, (3) AID, (4) AGH and (5) AFG.
Chiari-like malformation (CM) | |
Grade | |
No Chiari-like malformation | |
Cerebellum indented (not rounded) | |
Cerebellum impacted into, or herniated through, the foramen magnum. | |
Normal (no central canal dilation, no presyrinx, no syrinx) | |
central canal dilation or a separate syrinx | |
internal diameter of less than 2 mm or a pre-syrinx alone | |
central canal dilation which has an internal diameter of 2 mm or greater | |
a separate syrinx or pre-syrinx with central canal dilation | |
(measured as the maximum internal diameter in a transverse plane) | |
The syringomyelia grade is qualified with a letter indicating the age group | |
more than five years of age | |
three to five years | |
one to three years of age. |
All MRI measurements were made by author SPK who was blinded to the CM and syringomyelia status and the identity of the dog. Reliability of the MRI measurements was assessed as intra-measurer reliability using intraclass correlation coefficient (ICC) Model (2, 1)
The available data were used to develop discriminant functions for each analysis and then evaluated using a cross-validation technique thus avoiding data bias. Follow-up receiver operating characteristic (ROC) analysis on sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) was performed in an effort to assess clinical diagnostic relevance. The PPV and NPV in particular, allow the estimation of the proportions, respectively, of positives that are true positives and negatives that are true negatives. Microsoft Excel 2007 (
There was very high intra-measurer reliability for the line AE data, with an ICC of 0.985, 95% CI (0.944, 0.996) and for the F-diameter data with an ICC of 0.880, 95% CI (0.591, 0.969). The 155 dogs in the study ranged from 1.1 to 12.6 years old (mean 4.5 years, standard deviation 2.5 years; median 3.8 years). 53 (34%) dogs were under three years of age. 39 (25%) were 3.0- to 4.9 years 63 (41%) were 5 years and over. Five dogs had insufficient imaging of the forebrain to identify the optic nerve and optic canal and, for 11 dogs, the quality of the scan prevented the olfactory bulbs from being identified with certainty so that line AH and Angle 5 were not recorded for these dogs.
The distribution of CM and syringomyelia affectedness is depicted in
Number of dogs (%) | |||
CM | SM | ||
CM0 | 31 (20%) | SM0 | 35 (26%) |
CM1 | 32 (21%) | SM1 | 33 (24%) |
CM2 | 92 (59%) | SM2 | 67 (50%) |
Total dogs | 155 | 135 |
CM0 - no CM; CM1 - cerebellum indented by supraoccipital bone; CM2 - cerebellum impacted or herniated into foramen magnum; SM0 - normal; SM1 - central canal dilation or a separate syrinx, which has an internal diameter of less than 2 mm or a pre-syrinx alone; SM2 - syringomyelia (central canal dilation which has an internal diameter of 2 mm or greater, a separate syrinx, or pre-syrinx with central canal dilation). Syringomyelia may be late onset condition so SM0 dogs less than 3 years of age were excluded from statistical calculations for syringomyelia as their clear status could not be assured. Conversely SM1 and SM2 dogs less than 3 years old were included as identifying conformational and genetic indicators of this severe phenotype was a paramount goal of the study.
The yellow lines represent the significant variables in the study: F-diameter and Angle 2 and 5. Angle 3, and Lines AE and BC also found to be significant for CM and syringomyelia and are represented in the red (angle) and blue (lines). From left to right (i.e. Dogs A to D and E to H) there is an increase in CM and syringomyelia affectedness. The second and bottom rows detail the corresponding three dimensional images for each dog constructed from the MRI DICOM. For each dog, the age at time of MRI, maximum transverse width of the central canal dilation or syringomyelia is detailed. CCD – central canal dilatation; Yrs – years; CM – Chiari like malformation; SM – syringomyelia.
MRI measurements (mm) | |||||||
Dog (GriffonBruxellois) | CM | SM | F-diameter | AE | angle 2 | angle 3 | angle 5 |
A | 0 | 0 | 38 | 29.9 | 81.8 | 77.3 | 51 |
B | 1 | 0 | 42.6 | 31.7 | 76.9 | 81.2 | 46.8 |
C | 2 | 1 | 37.8 | 29.4 | 73.8 | 83.3 | 41.5 |
D | 2 | 2 | 45.6 | 33.1 | 63 | 83.7 | 44.7 |
E | 0 | 1 | 41.9 | 31.9 | 66 | 91.5 | 44.3 |
F | 0 | 2 | 42.8 | 34 | 66 | 76.6 | 41.7 |
G | 1 | 2 | 43.6 | 33.2 | 74 | 86.2 | 41.3 |
H | 2 | 2 | 46.8 | 34.7 | 62 | 103.9 | 42.2 |
I (Beagle) | 0 | 0 | 42.72 | 30.66 | 80.68 | 61.37 | 74.95 |
J (Aust. Terrier ) | 0 | 0 | 42.26 | 30.01 | 80.23 | 61.67 | 61.8 |
K (Aust. Terrier) | 1 | 0 | 39.54 | 29.31 | 72.4 | 63.78 | 62.5 |
L (Affenpinscher) | 0 | 0 | 40.85 | 30.95 | 69.63 | 88.04 | 45.75 |
M (Affenpinscher) | 0 | 0 | 41.08 | 31.11 | 69.42 | 69.02 | 45.17 |
N (Affenpinscher) | 1 | 0 | 41.32 | 30.96 | 60.29 | 89.08 | 46.5 |
Comparison between 8 representative Griffons (A–H) with different degrees of CM and syringomyelia affectedness (0,1,2) and also mesaticephalic and brachycephalic dogs with different degrees of CM affectedness (0,1) but without syringomyelia. Angle 2 becomes more acute with CM and Angle 5 becomes more acute with brachycephaly and this is irrespective of the height of the rostral cranial fossa (F-diameter). Aust. Terrier = Australian Terrier.
In
Top and third rows T1-weighted midsagittal MRI image with superimposed framework. The second and bottom rows detail the corresponding three dimensional images for each dog constructed from the MRI DICOM. Yellow: significant angles (Angle 2 and 5). Blue: significant lines (AE, BC, BD). Red: significant Angle 3. The mesaticephalic dogs L,M,N have a smaller angle 3 compared to the brachycephalic dogs I,J,K. Dog K has mild CM and has smaller angles 2 and 5 compared to normal mesaticephalic dogs. Likewise Dog N has mild CM and also has smaller angles 2 and 5 compared to normal brachycephalic dogs. CCD – central canal dilatation; Yrs – years; CM – Chiari like malformation; SM – syringomyelia.
The three groups, CM0 (No CM), CM1 (Mild CM) and CM2 (CM) comprised 31, 32 and 92 dogs respectively. Initial analysis of all possible variables indicated that 8 of the 15 variables were significantly different across the three groups (
Variable | F | p-value |
F-diameter | 9.468 | <0.001 |
BC | 7.975 | 0.001 |
BD | 3.828 | 0.024 |
AE | 5.394 | 0.006 |
FG | 4.439 | 0.014 |
Angle 2 | 10.552 | <0.001 |
Angle 3 | 4.299 | 0.015 |
Angle 5 | 4.883 | 0.009 |
F is the Analysis of Variance [ANOVA] test statistic. It assesses the ratio of between group variation to within group variation with higher values indicating the likelihood of a group effect.
Function 1 [p<0.001] D1 = − 0.351* F-diameter +0.112* Angle 2+6.968
Function 2 [p = 0.197] D2 = 0.423* F-diameter +0.118* Angle 2–26.357
The stepwise selection confirmed the two most significant variables from the ANOVA results (F-diameter and Angle 2) and the functions themselves are ordered by their level of statistical importance hence Function 1 is the more powerful. Interpreting Function 1 is based on how its value can be increased or decreased hence larger Angle 2 values accompanied by smaller F-diameters will increase this function (larger positive value). Conversely Function 1 will be smaller (more negative) when F-diameter is larger and Angle 2 is smaller. Function 2, whilst statistically non-significant also selected F-diameter and Angle 2 as the important variables and allows for some separation in another dimension, with larger measures of both variables increasing Function 2 values.
The group means (centroid; dark crosses) move more to the left as the CM0 to CM1 to CM2 i.e. CM absent to present.
SM0 and SM1 dogs over 3 years were combined to form syringomyelia “SM NO” group ensuring there was no bias for the late onset nature of syringomyelia and SM2 became syringomyelia “SM YES”. Initial analysis of the data from those 135 dogs across all possible variables indicated that 5 of the 15 variables were significantly different (
Variable | F | p-value |
F-diameter | 20.246 | <0.001 |
BC | 4.83 | 0.03 |
AE | 9.867 | 0.002 |
Angle 3 | 10.769 | 0.001 |
Angle 5 | 13.456 | <0.001 |
F is the Analysis of Variance [ANOVA] test statistic. It assesses the ratio of between group variation to within group variation with higher values indicating the likelihood of a group effect.
This confirmed the two most significant variables from the ANOVA results (F-diameter and Angle 5). Interpreting the function, as previously, is based on how its value can be increased or decreased hence larger F-diameters accompanied by a smaller Angle 5 will increase the value of the function (more positive) and conversely smaller F-diameters accompanied by a larger Angle 5 will decrease the value of the function (more negative).
The SM YES dogs have slightly higher values of the function (more positive) and the SM NO dogs have slightly lower values. A few SM YES dogs can be seen to be outliers (*) with unusually high function values.
The success of this methodology was assessed using cross-validation and yielded an overall correct classification result of 69.8%. This success was similar across the groups: SM NO 71.0% (44/62) and SM YES 68.9% (46/67). It was noted that six dogs were dropped from original 135 during the analysis as a result of missing data hence these results are based on 129 dogs.
The F-diameter, Angle 2 and Angle 5 variables have been shown to be the significant parameters when aiming to discriminant between the sub-categories of CM and syringomyelia but the power of the discriminatory functions varies considerably, both within a condition and between conditions (
Variable | No CM | CM | No SM | SM |
F-diameter | Smaller | Larger | Smaller | Larger |
Angle 2 | Larger | Smaller | – | – |
Angle 5 | – | – | Larger | Smaller |
‘Success’ | 61.30% | 54.30% | 71.00% | 68.70% |
The variables that yielded the most statistical power of discriminatory functions have been summarized as ‘success’. The magnitude of contribution from each variable is given (Smaller/Larger), with the level of “success” indicating the unbiased rate of dogs correctly classified. CM – Chiari like malformation; SM – Syringomyelia.
As F-diameter was the most significant variable in each of the previous discriminant analyses, the strength of this variable alone as a diagnostic tool was investigated further. For the 111 dogs where a clear indication of CM NO/CM YES [n = 28, n = 83] and SM NO/SM YES [n = 56, n = 55] was known, very significant statistical differences were found in the average length of F-diameter (p<0.001) for both CM and syringomyelia groupings. From inspection of mean differences a first estimate of an appropriate cut-off for having the condition (or not) was an F-diameter of 42.5 mm. Consequently this estimate was used with 0.1 increases and decreases as a starting point in assessing the sensitivity, specificity, PPV and NPV for both CM (n = 155) and SM (n = 135). ROC analyses produced optimum CM sensitivity of 70%, with specificity 71% and, for SM, a sensitivity of 77%, with specificity of 54%, both occurring at F-diameter of approximately 41.8 mm.
CM ROC – Area under curve = 0.715 (p<0.001), with estimated optimum cut for F-diameter 41.8 mm; Syringomyelia (SM) ROC – Area under curve = 0.695 (p<0.001), with estimated optimum cut for F-diameter 41.8 mm.
This study supports the multifactorial nature of CM and syringomyelia and that this condition cannot be explained by a simple defect in the development of a single skull bone; it is a more complex disorder involving cranial base shortening, craniocervical junction abnormalities and other, as yet undetermined, factors not investigated in this study. As an example, raised intracranial pressure and/or a mismatch in cerebrospinal fluid (CSF) production and absorption could influence development of syringomyelia but these parameters have yet to be investigated in detail
It is well recognized that brachycephalic dogs have a compensatory increase in cranial height to allow accommodation of the brain in the foreshortened skull
This dog has syringomyelia but is asymptomatic. To improve success in the show ring breeders may select for greater exaggeration of this characteristic which may increase the risk for symptomatic syringomyelia (picture courtesy of Lee Pieterse). Right: Young Griffon Bruxellois with CM and asymptomatic syringomyelia (picture courtesy of Henny van der Berg).
Line AE was also significantly greater for dogs with CM and syringomyelia. It is possible that this reflects increased height of the rostral part of the caudal cranial fossa (pars rostralis).
It has been shown previously that Cavalier King Charles spaniels with CM have a relatively large volume pars rostralis and relatively small volume pars caudalis
These figures were created using Adobe Photoshop (
Dogs with CM and syringomyelia were also more likely to have smaller Angles 2, 3 and 5. A more acute Angle 2 could result from a short basicranial axis in the caudal cranial fossa and (even more) premature fusion of the spheno-occipital synchondrosis. Angle 2 was uniquely significant to CM supporting this observation. Angle 2 could also be more acute if the atlas were closer to the occiput either because the supraoccipital bone had lost is convexity i.e. was less rounded or in the event of atlanto-occipital overlapping/proximity. Proximity of the atlas to the occiput would decrease the overall volume of the craniocervical junction; indeed, atlanto-occipital overlapping would not be a prerequisite for this to occur. Reduction in the volume of the craniocervical junction would decrease the size of the subarachnoid space and cisterna magna. This in turn could increase impedance to normal biphasic and pulsatile CSF flow resulting in inhomogeneous flow with abnormal CSF peak velocities predisposing to symptomatic CM and development of syringomyelia. However this hypothesis is not supported by the finding of a significant increase in length of line BC (basion of basioccipital bone to the cranial dorsal laminae of the atlas) which tended to be longer for both CM and syringomyelia groups. In addition Angle 3 tended to be smaller with CM and syringomyelia affectedness - a more acute Angle 3 would occur if the atlas was further away from occiput. However, both BC and Angle 3 was less useful for discriminating dogs with CM and syringomyelia and there are possible alternative explanations for the increase in line BC (
The conformation changes with CM and syringomyelia are summarized in
The T1-weighted midsagittal MRI image is from dog A (without CM or syringomyelia) and the framework of lines and angles is indicated in blue and higher case letters with the exception of angles 2 and 5 which are numbered in yellow. The framework of dog D (with CM and syringomyelia) has been superimposed on the image and aligned with baseline HAI and on the F-diameter. The framework of dog D is in red and with lower case letters. Angles 2 and 5 are yellow. It is possible to appreciate how with syringomyelia, the occipital lobe circle and height of cranial fossa increases (red circle). Angles 2 and 5 (yellow) are decreased as a consequence of the cranial base shortening and increased proximity of the atlas to the occiput. In addition the vertexes at E and D (red triangles) are closer together giving an appreciation of the overcrowding and change in shape of the caudal fossa. The white arrows depict the changes between the measurements.
One of the aims of this study was to develop a system of measurements that could be used on MRI obtained by machines of differing field strength and which do not include the entire forebrain. This is necessary because for “low cost” MRI screening a limited but economic imaging is obtained to ascertain CM and syringomyelia status prior to breeding
Angles proved more useful than point-to-point distant measurements because the latter are affected by head and body size, which is variable even within the same breed. For example we found that the standard error of the caudal skull base length (AB) was almost as large as the actual mean difference hence it was a non-significant. By contrast angles may represent a more accurate representation of the position of one anatomical feature relative to another. However a possible weakness of this study is that the size, weight and body condition of the dogs was unknown and sexual dimorphism was not analyzed which might have affected some parameters.
This study achieved its aim which was to establish a method of qualitative analysis of MRI DICOM images to enable better phenotyping and therefore genotyping. However, it failed to identify a sensitive and specific measurement(s) to improve the current British Veterinary Association and Kennel Club MRI health scheme and better diagnose CM and the risk of syringomyelia at an early age thus precluding the need for dogs to undergo multiple MRI evaluations throughout life. This is most likely because diseases like CM and syringomyelia have a complex etiology and disease expression probably occurs when there is a threshold combination of genetic and environmental factors. Therefore further work is required and we recommend that this study be repeated and/or modified for other breeds with a high prevalence of CM and syringomyelia such as Cavalier King Charles spaniels and Chihuahuas.
This study supports the view that CM is a multifactorial condition that includes the shortening of the entire basicranium, loss of convexity of the supraoccipital bone, invagination of the cerebellum under the occipital lobes and possibly by increased proximity of the atlas to the occiput. As a compensatory change, there is increased height of the rostral cranial cavity and lengthening of the dorsal cranial vault. Overcrowding in the caudal cranial fossa and the craniocervical junction is a defining feature. The study provides the basis of a quantitative assessment of CM which might identify risk of syringomyelia and suggests that CM should be redefined so that account is taken of the overcrowding of the entire cranial fossa and craniocervical junction with reorganization of the brain.
(XLSX)
(XLSX)
The study was made possible by a worldwide group of breeders and pet owners who participated in and supported the research into CM and syringomyelia. Particular thanks are given to Lee Pieterse, Henny van der Berg, Jessica Gruninger, Rachael Harvey, Sandy Smith, the nursing staff at the Stone Lion Veterinary Hospital and the many radiographers and veterinary surgeons who imaged the dogs particularly those in Australia, Netherlands, Finland and the Bioimaging Research Center University of Georgia (USA).