Peri‐ictal magnetic resonance imaging characteristics in dogs with suspected idiopathic epilepsy

Abstract Background The pathophysiology of changes in magnetic resonance imaging (MRI) detected after a seizure is not fully understood. Objective To characterize and describe seizure‐induced changes detected by MRI. Animals Eighty‐one client‐owned dogs diagnosed with idiopathic epilepsy. Methods Data collected retrospectively from medical records and included anatomical areas affected, T1‐, T2‐weighted and T2‐FLAIR (fluid‐attenuated inversion recovery) appearance, whether changes were unilateral or bilateral, symmetry, contrast enhancement, mass effect, and, gray and white matter distribution. Diffusion‐ and perfusion weighted maps were evaluated, if available. Results Seizure‐induced changes were T2‐hyperintense with no suppression of signal on FLAIR. Lesions were T1‐isointense (55/81) or hypointense (26/81), local mass effect (23/81) and contrast enhancement (12/81). The majority of changes were bilateral (71/81) and symmetrical (69/71). The most common areas affected were the hippocampus (39/81) cingulate gyrus (33/81), hippocampus and piriform lobes (32/81). Distribution analysis suggested concurrence between cingulate gyrus and pulvinar thalamic nuclei, the cingulate gyrus and parahippocampal gyrus, hippocampus and piriform lobe, and, hippocampus and parahippocampal gyrus. Diffusion (DWI) characteristics were a mixed‐pattern of restricted, facilitated, and normal diffusion. Perfusion (PWI) showed either hypoperfusion (6/9) or hyperperfusion (3/9). Conclusions and Clinical Importance More areas, than previously reported, have been identified that could incur seizure‐induced changes. Similar to human literature, DWI and PWI changes have been identified that could reflect the underlying metabolic and vascular changes.


| INTRODUCTION
Magnetic resonance imaging (MRI) can identify seizure-induced changes in the ictal and early postictal stages of seizure activity. 1 The pathophysiology responsible for these changes is not fully understood. Seizure activity increases regional glucose and oxygen consumption and utilization. It has been proposed that to compensate for this increase in metabolic demand, compensatory regional hyperperfusion occurs. 2 However, should sustained ictal activity persist, these compensatory mechanisms are no longer sufficient. This leads to hypoxia, lactic acidosis and failure of cellular homeostasis causing cellular swelling and increased membrane permeability. The cascade of metabolic changes subsequently leads to the formation of cytotoxic and vasogenic edema, a finding supported by experimentally induced seizures in rats. 3 The formation of identifiable changes has been documented as time-dependent 4 ; especially dependent on duration of ictal activity.
Peri-ictal changes are well-documented in human literature. The prevailing conclusions from such investigations 1 were that these changes occur in specific anatomical areas of the brain, they can be unilateral or bilateral, local or remote to the epileptogenic focus, and are partially or completely reversible; resulting in partial or complete resolution. Diffusion and perfusion-weighted imaging support the presence of pathophysiological changes within the ictal tissue and suggest a timedependent evolution of these changes. The reversible nature and anatomical locations are described in 3 dogs with seizures of different etiologies, 5 as well as the reversible nature of these changes.
The aims of our study were to further describe the changes and anatomical locations seen on MRI during the peri-ictal stages in a group of dogs diagnosed with presumed idiopathic epilepsy and describe the anatomical locations affected. If available, we aimed to report if any changes were seen on diffusion-and perfusion-weighted imaging. Medical records were searched for dogs presenting for investigation of epileptic seizures and that had peri-ictal changes identified on MRI of the brain between 2010 and 2019. The terms "postictal" and "peri-ictal" were searched within the imaging reports and the studies were identified. The inclusion criteria were further refined to include dogs that were diagnosed with suspected idiopathic epilepsy. The criteria included dogs that had an onset of seizures between 6 months and 6 years, had a normal cerebrospinal fluid (CSF) analysis, and did not develop progressive neurological deficits after 6 months after diagnosis. Cases were also included if repeated CSF analysis was normal, after an abnormal initial result and only if naive of medication that could influence CSF analysis. Further data collected included age, breed, sex, type and frequency of the most common seizure activity, and the duration of time between the last seizure and acquisition of MRIs. Exclusion criteria involved MRI abnormalities that were not consistent with reports in human and veterinary literature.

| Magnetic resonance imaging
All the images were obtained from the following scanners: 1. All the images were reviewed by a board-certified radiologist (FMc).
The imaging data obtained included anatomical areas affected, T1-weighted, T2-weighted and T2-FLAIR (fluid-attenuated inversion recovery) signal characteristics, whether the changes were unilateral or bilateral, symmetrical or asymmetrical, presence of contrast enhancement, mass effect and gray and white matter distribution. For those that had diffusion-weighted imaging, the affected areas on conventional imaging were assessed. Diffusion weighted imaging (B = 0 and B = 800 or 1000) and apparent diffusion coefficient (ADC) maps were used for analysis. Similarly, for those that underwent perfusionweighted imaging, changes were assessed based on the affected areas on conventional MR imaging. Relative cerebral blood flow (rCBF), relative cerebral blood volume (rCBV) and mean transit time (MTT) maps were needed for interpretation. Quantitative values were obtained from the ADC, rCBF, rCBV, and MTT maps. They were obtained by manually drawing up to 3 regions of interest (ROIs) in each affected area identified on conventional imaging (T2-FLAIR) and formulating a mean value. Care was taken to exclude large blood vessels from regions of interest.
Two separate control groups were also formulated: one to compare diffusion-weighted changes and another to compare perfusionweighted changes. These control groups were composed of dogs diagnosed with primary myopathies or cranial nerve neuropathies without central nervous system involvement. Only images where lesions were free from artifacts were used for analysis. Quantitative ADC and perfusion map values (rCBF, rCBV and MTT) were obtained, from the respective control groups, of the areas that were known to be affected in the study groups. The median measurement values were subsequently obtained for each affected area based on each study group.
To assess whether the affected area had facilitated diffusivity, restricted diffusivity or normal diffusivity, the mean ADC values of the study group were compared to values obtained from manually drawn ROIs of the unaffected contralateral hemisphere (if unilateral changes) or compared to median values of the control group of dogs (if bilateral); that is, a higher comparative ADC value implied a facilitated diffusivity and a lower ADC value implied restrictive diffusivity.
A similar method was applied to the mean perfusion measurements to assess whether an area showed evidence of hyperperfusion or hypoperfusion; that is, hyperperfusion was identified as increased rCBF, increased rCBV and reduced MTT, and hypoperfusion was identified as decreased rCBF, decreased rCBV, and prolonged MTT. Regions of interest in both diffusion and perfusion weighted imaging maps were defined as abnormal if the quantitative values were greater than 2 standard deviations of the control or contralateral hemisphere.

| Statistical analysis
Coexisting or concurrent peri-ictal changes were defined as identification of 2 or more lesions in different brain regions within the same animal. A concurrent lesion network was created using graphical network analytical methods, implemented in the igraph package available in R. 6 A node referred to a brain location, with the size of the node being relative to the frequency of lesions at that location. An edge (link) referred to a coexisting lesion at a different location, with edge weight being relative to frequency of coexisting lesions. This undirected connected graph (network) consisting of 10 nodes and 35 edges was then examined for group structure via an optimal community structure algorithm, with the aim of defining groups (communities) of commonly seen concurrent lesions. Briefly, this algorithm partitions the network structure to maximize its modularity function (Q) and identify the optimal community configuration consisting F I G U R E 1 Magnetic Resonance transverse images of 1 case at the level of the mesencephalon; A, T2-weighted, B, T2-weighted fluidattenuated inversion recovery (FLAIR), C, T1-weighted precontrast and D, T1-weighted postcontrast. These images highlight T1-hypointense, T2-hyperintense areas affecting the cingulate gyrus (encircled area) and the hippocampus. These hyperintensities do not suppress on FLAIR. Contrast enhancement is identified in the periphery of the affected areas including the meninges (white arrows) and the ependymal border of groups with many more internal connections than expected at random. 7 A Fisher's Exact test was performed to assess whether there was significant association between the diffusion and perfusion-weighted findings and was performed on a standard statistical software package (SPSS:  D); at the level of the caudal diencephalon. The T2-hyperintense areas appear to have very high signal intensity affecting the gray matter (white arrowhead) and mildly higher signal intensity affecting the white matter (black arrows) on the DWI image. The mildly increased signal affecting the white matter could be due to T2-shine through phenomenon. On the ADC map, there is high signal within the white matter (black arrowhead) and low signal affecting the gray matter (white arrow). This could indicate facilitated diffusion in the white matter and restrictive diffusion in the gray matter of the cingulate gyri performed 1 week after the initial MRI and in 1 dog after 10 months.
Full resolution of changes was identified in the latter dog and in one of the early scans ( Figure 3). In the other MRI performed 1 week after initial investigations, there was a marked reduction in the size of the original changes. All 3 of the repeated scans had abnormal initial CSF analysis, namely mild increase in total nucleated cell concentration and protein concentration (Supporting Information File S1). On repeated analysis, all 3 cases had normal CSF analysis.

| Distribution data
Distribution analysis was carried out to identify if there was an association between certain anatomical areas. The results reveal that multiple affected locations were more often found than just a single location ( Figure 6). This revealed the following relationships: 80% and 89% of the all cases of cingulum lesions observed were present when a parahippocampal lesion and pulvinar thalamic lesion was identified respectively, and 70% of all cases of hippocampal lesions observed were seen in the presence of a parahippocampal lesion. Connectivity of certain locations was seen ( Figure 7) and imply that there is increased connectivity between olfactory lobe, caudate nuclei and frontal, between the piriform lobe and the hippocampus, and between the cingulate gyrus, parahippocampal gyrus, occipital lobe, pulvinar thalamic nuclei, and the temporal lobe.

| DISCUSSION
The most common acute peri-ictal MR changes include bilateral symmetrical T2-and T2-FLAIR hyperintense, T1-hypointense area with mild contrast uptake and have tendency for local mild mass effect. The most common sites affected included the hippocampus, cingulate gyrus, and piriform lobes; although other locations have now been identified. Abnormalities in the diffusion and perfusion maps have also been recognized in this study; which might reflect the underlying pathophysiological processes.
The imaging abnormalities identified on conventional MRI in this study, are similar to those reported elsewhere. 5 Studies in rodents have identified that the T2-weighted hyperintense areas are a result of ictal-related edema or early or developing gliosis. 1,8 The patterning of peri-ictal gadolinium enhancement seen correlated well with what has been previously reported in human and animal studies; with a discrete area of leptomeningeal enhancement around the affected T2-hyperintense areas most commonly identified. 5,9 The proposed mechanism for this is that prolonged ictal activity leads to local lactic acidosis, an increase in carbon dioxide and neurotransmitter release, leading to an increase in local vascular permeability. 9 There are currently no studies in humans or in domestic species correlating seizure-induced changes with neuropathological findings.
One hypothesis behind the increased T2 signal could be related to swelling or hypertrophy of glial cells leading to prolonged T2-relaxation times. This is based on theories on dysregulation of aquaporin-4 channels 10 and failure of Na/K pumps 11 leading to the formation of vasogenic and cytotoxic edema. 3 Conventional imaging cannot distinguish between cytotoxic and vasogenic edema, thus more advanced techniques such as diffusionweighted imaging have been used. Diffusion weighted imaging assesses the movement and distribution of water molecules within a given time within a specific location. Changes in signal on DWI and on ADC maps in dogs with kainite-induced status epilepticus are caused by ictal activity. 12 A reduction in ADC corresponds with decreased diffusivity which could be directly associated with cytotoxic intracellular edema due to intracellular ictal-driven excitotoxic mechanisms.
Other theories for a decrease in ADC values include macrophagic, microglial, and astrocytic proliferation. 3 An increase in ADC values is commonly associated with vasogenic edema formation or progressive neuronal loss. Studies in humans have targeted their efforts to evaluate the fluctuations in ADC values as they are thought to reflect neuronal metabolic dynamics associated with seizure activity. 13,14 Some of these studies identify a reduction in ADC values; while others have identified a transient increase in ADC values within the peri-ictal and interictal period. 3,[12][13][14][15] The variability of findings could be timedependent and follow certain stages; stage 1 initial regional hyperperfusion with no detectable ADC change, stage 2: high ADC values due to vasogenic edema formation, stage 3: reduction in ADC values resulting from cytotoxic edema formation, and, stage 4: increased ADC values due to progressive neuronal loss. 16 In idiopathic epileptic dogs, ADC values have been shown to be increased in the amygdala within the interictal period 17 and it was speculated that this change could be secondary to neuronal cell loss or gliosis. The results of this study similarly identified a variety of patterns of diffusivity. The small size of the group limits statistical power. However, the variation might suggest a similar dynamic time-dependent change to the ictal and paraictal tissues. In some cases, both high and low signal were identified on the ADC maps, suggesting that these stages might overlap. Consistent changes in the piriform lobes and the amygdala are similar to interictal findings in idiopathic epileptic dogs. 17 A subgroup of our cases had a normal signal on ADC maps but an increased signal on DWI. This could be associated with different underlying mechanisms including that DWI is not only dependent on ADC map but also on the T2-weighted appearance. Therefore, having a high signal on DWI and a normal signal on ADC could imply "T2-shine through". Other possibilities for a normal signal on an ADC map are that there could be combined effects of vasogenic and cytotoxic edema, or, cytotoxic edema and gliosis, affecting the same area of tissue. Lastly, it is possible that the images acquired in the 3-Tesla scanner, compared to the 1.5-Tesla, gave a better resolution of images, making it easier to differentiate between areas of reduced and increased signal. It is well established in humans that there is an initial local hyperperfusion during ictal activity followed by hypoperfusion in the postictal period. 18 Our study group has identified that these changes in perfusion occur in dogs.
The affected anatomical areas identified are more extensive than those previously reported. 5 The cingulate gyrus in particular features as one of the most prevalent areas affected, alongside the piriform lobe and the hippocampus. Another new location identified in this study was the medial pulvinar thalamic nucleus. It was interesting to note that all the thalamic affected areas were in this particular location. Electrophysiologic and resting state functional connectivity human studies demonstrate that the medial pulvinar nucleus is particularly important in genesis and reciprocal propagation of seizure activity with the mesial temporal structures. 19,20 This study identified that some affected areas had potential influence over the likelihood of other areas being affected. One possibility is that two locations could have a contiguous relationship. Examples of this association include the cingulate gyrus and parahippocampal gyrus and the parahippocampal gyrus and hippocampus, which have been identified in our study as frequently affected at the same time.
Another suggestion is that there likely are projection fibers connecting certain other areas with each other. An example of this would the pulvinar thalamic nuclei and certain areas of the cortex due to thalamocortical relay circuits. The anatomical areas identified, and the connectivity proposed, resembles an already defined circuitry; the Papez's circuit. The Papez's circuit is a fundamental component of the limbic system and therefore a control of emotional expression and memory. The structures include the entorhinal cortex, hippocampus, fornix, mammillary bodies, parahippocampal gyrus, cingulate gyrus, amygdala, and anterior thalamus. Abnormalities of Papez's circuitry has been identified in human patients with mesial temporal epilepsy. 21 CSF abnormalities were identified in 3 cases, which had normalization of measurements on repeat analysis within a few days without specific treatment. Increased protein and total nucleated cell concentration are a transient phenomena in humans and suspected in dogs although this finding remains controversial. [22][23][24] As part of the inclusion of this study, we identified two cases that had bilateral necrotic changes to the amygdala and piriform lobes.
These changes have been documented in experimental studies as well as other retrospective imaging-based studies. 12,25 However, these have only been described in the interictal period of idiopathic epileptic dogs. In addition to these studies, hippocampal atrophy has also been observed in dogs. [26][27][28] Such peri-ictal changes might be the prequel to such irreversible changes; brain atrophy and cortical laminar necrosis, as a result of sustained epileptic discharges in humans. 29 There are several limitations in this study, mostly related to its retrospective nature. There were few repeated studies (only performed in 3 cases) and for this reason it was difficult to completely exclude other differentials for seizure activity such as metabolic, vascular, or neoplastic changes. Also, without this information, pertinent data related to evidence of persistent changes and whether they were related to the areas of suspected cytotoxic edema could not be evaluated. The perfusion maps were based on T2* based sequences and some of the DWI were also gradient echo-based. This results in susceptibility artifacts around air-filled chambers such as the frontal sinuses and tympanic bullae. In some cases, the artifacts distorted the areas of interest making it difficult to analyze and interpret. Lastly, it was difficult to fully define the boundaries between piriform cortex and amygdala on conventional imaging; due to close proximity of the structures and similarities in appearance in T1-and T2-weighted sequences. Volumetry using higher resolution MRI and 3-dimensional software would need to be incorporated to analyze amygdaloid changes in more detail. 30 The variability of the MRI changes in this study implies limited clinical utility. These abnormalities can be a mixture of local and remote changes and thus adjunctive diagnostic utility of electroencephalography should be used to aid the diagnosis of the epileptogenic focus. 31 Ischemic infarcts can be difficult to differentiate from the MRI changes described in this study. However, the combination of hyperperfusion, increased T2-signal and restricted diffusion is uncharacteristic of ischemia. 32 Also, ictal changes will not align vascular territories. 33

ACKNOWLEDGMENT
No funding was received for this study. This work has been presented at the 32nd ECVN/ESVN Annual Symposium in Wroclaw 2019.

CONFLICT OF INTEREST DECLARATION
Authors declare no conflict of interest.

OFF-LABEL ANTIMICROBIAL DECLARATION
Authors declare no off-label use of antimicrobials.

INSTITUTIONAL ANIMAL CARE AND USE COMMITTEE (IACUC) OR OTHER APPROVAL DECLARATION
Authors declare no IACUC or other approval was needed.