To the author’s knowledge this is the first study to use Cavalieri’s stereological principle to estimate brain volume in dogs with IE. In this study, no differences in FB, SAS and LV volume were found between dogs with IE and controls. Associations between clinical history and brain volume were found within the IE group but, as will be discussed further below, these results require further research to substantiate.
Effect of IE on FB, SAS and LV volume
This study found no statistically significant differences in FB, SAS and LV volume in dogs with IE compared to controls. Whilst many studies have focused on LV volume, we believe this is the first study to report FB and SAS volumes in dogs with IE compared to controls. Unfortunately, “negative” results such as these can be easily dismissed, but these results further support the need for a rigorous diagnosis of exclusion for dogs with IE, given that there are still no diagnostic biomarkers specific to IE. Furthermore, volumetric studies are important in veterinary medicine to establish a database of “normal” volumes for various breeds and disorders.
Most volumetric studies in dogs with IE compared to controls focus on total LV volume (as in this study) or on left/right LV asymmetry [13, 21, 24, 27, 28] reporting mixed findings. This study contributes to this existing body of research, perhaps strengthening the conclusion that there is no difference between total LV volume in dogs with IE compared to controls. Similarly, volumetric studies in people with epilepsy have reported various additional structural abnormalities [4,5,6,7,8,9, 11, 13], though a unanimous hypothesis on the cause of these changes has not been reached.
Historical veterinary studies [20, 25, 30] have suggested normalising the volume of the LV to the overall brain volume, especially when comparing different breeds. Using the data acquired in this study we expressed FB volume as a relative value and expressed LV volume as a percentage of FB volume, but results did not change the conclusion already made. Importantly, the body weights of dogs included in this study were not significantly different between IE and control groups, as Schmidt, et al. [31] have shown that brain volume is significantly affected by body weight in dogs. Though, it is worth mentioning that body condition score was not taken into account by either Schmidt, et al. [31] or the present study.
Pilegaard, et al. [32] found that breeds with a wide (laterolateral) but shallow (rostrocaudal) skull shape had a smaller cortical volume to ventricular volume ratio. Similarly, Bakici, et al. (2019) [33] used the Cavalieri method to compare brain fraction volumes, including lateral ventricles, of brachycephalic and mesocephalic breed types and found no significant difference, although the aforementioned authors’ findings should be interpreted with caution because, in stereology, volume fractions are normally limited and sometimes inconclusive and cannot be directly compared with more accurate global volumes as the ones we estimated in our study. Of the eight breeds included in this study, all are mesocephalic aside the Boxer, which was not measured due to low numbers. Therefore, the skull type of the dogs included is unlikely to have skewed the data as all dogs were a similar size and skull shape. Additional analysis to confirm this assumption by comparing FB, SAS and LV volume for Border Collies, Labradors, German Shepherd Dogs and Golden Retrievers with IE to controls separately by breed showed no significant differences in volumes for the FB, SAS or LV in any of the breeds, which further supports the negative findings reported.
Effect of time since the onset of IE
The time since the onset of IE was moderately-positively correlated to volume of the LV, but this finding was only a statistical trend. Figure 2 suggests this result is largely driven by two dogs, a Labrador Retriever and Golden Retriever, with large LVs where onset of IE was more than 1000 days prior to MRI. To assess this, a multivariate analysis of variance (MANOVA) was carried out, which showed no significant interactions between breed, LV volumes and time since seizure onset. Given the dataset was small, the dataset was copied once to double it. This second analysis with the doubled dataset showed a significant relationship between breed, LV volume and time since seizure onset (p = 0.005), of which the significant contributing interaction was breed and LV volume (p = 0.001). In this instance breed explained 34.8% of the LV volume difference. A follow-up one-way ANOVA showed Golden Retrievers to have significantly larger LV volumes than Border Collies and German Shepherd Dogs. Based on these rudimentary results, the authors would suggest careful selection and control of dog breed in the further longitudinal analysis of LV volume in dogs with IE recommended above. Though tempting to include many different dog breeds in a study to broaden the application of findings, the large conformational diversity of dogs may prohibit this by introducing avoidable variability.
A relationship between these variables in dogs with IE is biologically plausible; in a study on rat models of SE [34], it was found that LV volume increased over time. However, in dogs with IE, Kuwabara, et al. [28] reported no association between the hippocampi asymmetry ratio and the time since onset of seizures.
It has previously been reported that LV dilatation is positively correlated to increasing age in dogs [35,36,37] and people [38] and, therefore, age could be a confounding variable, in addition to the breed factors already discussed. Furthermore, Noh, et al. [39] have recently shown brain atrophy in a group of dogs over 9 years of age with cognitive dysfunction and, therefore, the effect of age on this population must be considered. The maximum age of dogs in this analysis was 79 months (6 years and 7 months), which would be considered middle-age. However, a previous study investigating canine cognitive dysfunction (CCD) in dogs with IE found onset of CCD-related clinical signs to be earlier than expected [40]. Without the support of MRI scans we cannot know if this is due to structural changes, but it would be an excellent route for further study. Dogs less than 6 months of age were not included, per the lower limit in the IVEFT guidelines for diagnosis of IE [3].
Though strong conclusions cannot be made, this study suggests that LV volume change over time could be a valuable future research hypothesis to test in a large veterinary longitudinal MRI study.
Cluster seizures and status Epilepticus
Whilst it is likely that dogs with CS or SE have been included in groups of dogs that have IE during volumetric studies, to the authors’ knowledge they have never been assessed as a separate entity. The FBs of dogs with SE were found to be significantly larger than controls, which was surprising and did not match clinical suspicions held prior to the study. Further investigation showed poor breed distribution in these two groups despite systematic and uniform random sampling (SURS); the SE group contained only Labradors and Golden Retrievers, whilst the control group only contained one of these dog breeds. Additionally, an outlier in the control group shows a Border Collie with a particularly small forebrain. It is likely that breed distribution is a factor in this result too and, therefore, we cannot confidently conclude that the FB is larger in dogs with IE who experience SE. Despite following standard methodology for stereological studies with regards to group sizes, we recommend breed-specific or larger breed-diverse analysis is carried out in future to identify significant trends should they exist..
Despite our results not aligning with expectations and likely breed-bias of our results, post-ictal changes in brain structure following a seizure or SE have been reported in people [41] and dogs [3]. In human medicine it is known that prolonged seizure activity causes neuronal death [42], and Hocker, et al. [43] demonstrated that cortical and subcortical atrophy occurs in people who experience super-refractory SE (SE that recurs or continues for more than 24 h despite the use of anaesthesia). Newey, et al. [44] reported one case where global cerebral atrophy occurred progressively in a patient with super-refractory SE.
Use of the Cavalieri principle
The Cavalieri principle was easy to use after a period of training and practice to familiarise the observer with the software, however, it was a time-consuming process with each brain taking 60–90 min to measure. The high density of cross-point markers was necessary to account for the relatively small volume of the LV so that a sufficient number of points could be counted to ensure an acceptable CE, meaning that thousands of cross-points were counted in larger areas such as the FB. If an observer was only interested in one anatomical structure then the density of the cross-points could be adjusted to account for this, thus speeding up the process.
Two studies have detailed the accuracy of the Cavalieri principle using MRI scans compared to the actual brain post-mortem; Furlong, et al. [45] found poor agreement between MRI and pathological volumetric analysis of the cortical and subcortical volume and thickness of the cerebral cortex in human brains post-mortem, and they concluded that this was likely due to the poor resolution provided by the 3 T MRI used. A similar study [46] used the Cavalieri principle to compare pig brain volumes on 3 T MRI under anaesthesia and on physical sections following immediate euthanasia. They found no statistically significant differences between estimation methods in all areas except the basal nuclei compartment. Whilst this is not considered an appropriate statistical measure for agreement, the authors reported mean differences of 3–5% for the total brain volume, which are acceptable, but up to 9–11% for the other compartments, indicating a less satisfactory level of agreement than the t-tests suggest. Given this evidence it could be that smaller and more subtly defined areas measured using the Cavalieri principle on MRI do not agree well with true physical sections due to poor resolution of the anatomical structure represented in the MRI or that a finer grid needs to be used. The use of contrast solutions (e.g. gadolinium) during MRI could be investigated as a potential way to improve the resolution required for the better visualisation of the anatomical area of interest.
One challenge of using the Cavalieri principle with digital images can be seen in Fig. 4. Due to the nature of MRI images representing the average of a slice thickness, potential for a partial volume effect, the use of a fine saturation gradient of white, grey and black, and the need to magnify an image for accuracy thus creating a very pixelated image, means measurements can have a certain degree of subjectivity.
Limitations
The presence of two dogs with mildly increased protein in their CSF presumed due to recent seizure activity could have biased the study. The IVEFT recommend repeating these tests 16 and six weeks later, respectively [3], however, in veterinary medicine this is unlikely to occur because care is privately funded. For further support of IE diagnosis follow-up clinical examination can be repeated 1–3 years later to confirm no development of interictal neurological deficits. Due to the referral population of these dogs the mean days since seizure was well under this recommendation.
The MRI scanners used in this study were 1.5 T, therefore, not comparable to 3 T and above scanners available in human medicine. It is possible that with a 3 T scanner, differences could be found or that longer scans on the 1.5 T scanners would be preferable for volumetric analysis. A study comparing image quality, resolution and contrast of MRI images of the canine brain taken on a 3 T and 7 T MRI scanner found that both had their preferred uses and that some structures were better visualised on 3 T and some on 7 T [47]. The authors suspected that the use of 7 T MRI could help in the explanation of certain pathology in brain disorders such as epilepsy, but access to these powerful magnets is very limited in a veterinary clinical or research environment, which inhibits further research in this area. T2 TRA sequences were chosen because they offered good tissue contrast and were commonly used across patients with epilepsy and controls, and across both hospital sites. This decision has potential to affect the measurement process and should be taken into consideration when comparing to the findings of existing or future research.