Background

Canine epilepsy has an estimated prevalence of 0.62–0.75 % in primary veterinary practice [1, 2] and as such is one of the most common chronic neurological diseases. Magnetic resonance imaging (MRI) is regarded as an essential diagnostic test however the specificity is limited because the diagnosis of idiopathic epilepsy is one of exclusion and the reliability of diagnosis is limited by available technology and expertise in interpretation. The International League against Epilepsy (ILAE) defines idiopathic epilepsy as an epilepsy of predominately genetic or presumed genetic origin and in which there is no gross neuroanatomic or neuropathologic abnormality [3]. Therefore by default, MRI examination of an animal with idiopathic epilepsy should be “normal” (in human epilepsy termed MRI–negative). However the ability to detect lesions depends on many factors that affect the quality of the MRI examination (Table 1). Some of these factors can be controlled, such as optimal slice thickness and sequence. Other factors are less easy to influence. For example, the ideal epilepsy protocol in humans (Table 2) would include a gradient echo or similar technique for detecting haemorrhage or calcification. However this sequence is sensitive to susceptibility artefacts arising from the skull bones for example the mastoid area of the temporal bone, which are a more significant problem in veterinary patients that have a greater bone:brain ratio than humans. The interface between bone and air can cause inhomogeneity in the magnetic field and signal void (susceptibility) artefact, particularly noticeable on special sequences such as diffusion-weighted imaging (DWI) and which can interfere with MR spectroscopic techniques.

Table 1 Factors that have an effect on the ability to detect epileptic lesions on MRI
Full size table
Table 2 Epilepsy-specific MRI protocol for humans This “essential” 6 sequence protocol allows the detection of virtually all common epileptogenic lesion in humans and was proposed after systemic analysis of 2740 patients in a epilepsy pre-surgery program [13, 38, 41]
Full size table

The ability to detect epileptogenic lesions is further limited by economics. For example, imaging with a 3 tesla (3 T) MRI system gives better anatomical detail and is superior for detecting subtle lesions such as mesial temporal sclerosis [4] and migration disorders [5, 6]. However the initial and on-going cost of this technology is prohibitive for many institutions and indeed much of veterinary MRI is performed on low field (1 T or less) scanners, which have decreased spatial resolution and signal-to-noise ratio (SNR) [7].

Other technology may need to be employed to detect lesions in MRI–negative patients. Methods of processing MRI data post-acquisition have identified previously undetectable or overlooked abnormalities in humans [8, 9]. One such example is employed to improve hippocampal volumetric measurements in the sparsely myelinated and small brain of neonatal humans. To achieve this, contrast is optimised by combining dual echo T2W and proton density images [10]. In large part this is based upon the fact that discovery of a surgically resectable lesion significantly improves the prognosis in human drug-resistant focal epilepsy, including abnormalities of the hippocampus in the region of the mesial temporal lobe. As a result, if the MRI is negative then further work-up, for example with functional MR imaging, is engaged to help localize the epileptogenic lesion [11–13]. Table 3 details examples of the modalities used, none of which are established as routine in animals. However before making recommendations for advanced imaging, the veterinary surgeon and the owner must be clear about what is to be gained. Unless the diagnostic procedure changes the outcome or management there may be little achieved by subjecting an animal to invasive and/or expensive procedures. For example, Smith and others found that if an epileptic dog was less than six years old and had a normal inter-ictal neurological examination then there was a 97 % confidence of a unremarkable low field brain MRI, making diagnosis of idiopathic epilepsy very likely [14]. At present, given the lack of surgical or other therapeutic techniques available to improve prognosis over standard antiepileptic therapy, more research is required to improve the diagnostic sensitivity of MRI and establish the value of such therapeutic techniques.

Table 3 Novel imaging modalities for identifying epileptic foci
Full size table

The purpose of this article is to propose an epilepsy-specific MRI protocol that will optimise detection of lesions ruling out idiopathic epilepsy as a diagnosis, standardise the diagnosis for entry into clinical trials and facilitate detection of lesions which develop as a consequence of epilepsy, as well as provide high quality data for future studies investigating the pathophysiology of epilepsy.

Aim of advanced diagnostic imaging for animals with epilepsy

There are three main aims of advanced diagnostic imaging of the epileptic animal: 1) to rule out causes of epileptic seizures which may be treatable with means other than antiepileptic therapy only (e.g. inflammatory or infectious brain disease) 2) to identify lesions which are caused by seizures but are not themselves the source of seizures for example hippocampal sclerosis and 3) to provide data to further advance the field of research into the pathogenesis and/or treatment of epilepsy. Importantly, MRI must always be preceded by a thorough investigation including a good clinical history with clinical and neurological examination (see Consensus Proposal on the diagnostic approach to epilepsy in dogs). In addition, the absence of lesions identifiable on MRI examination does not indicate prognosis or which drugs are most appropriate. However MRI may enable the detection of lesions that may be associated with drug-resistance such as hippocampal sclerosis [5]. High resolution imaging of the hippocampus is therefore paramount in humans but the value of this remains undetermined in animals [15, 16].

Identification of the epileptogenic lesion

Most veterinary hospitals that offer advanced diagnostic imaging use the same protocol for the epileptic brain as for detection of gross intracranial pathology such as tumours. This reflects the aim of the procedure, namely to identify those lesions that have a different prognosis or treatment to idiopathic epilepsy. In human medicine, different MRI protocols are performed depending on whether the patient is expected to have idiopathic or structural epilepsy. Some might recommend that epileptic animals that are not expected to have idiopathic epilepsy (for example those animals younger than 6 months or older than 6 years or those patients with abnormal inter-ictal neurological examination) could be examined using an MRI protocol that does not require as high a resolution imaging of the brain while those patients expected to have idiopathic epilepsy could be examined using a higher resolution protocol. However in practice the expense and risk associated with general anaesthesia in veterinary patients makes it unlikely that more than one protocol be used for scanning an animal with epileptic seizures. Therefore any protocol developed for animals must be capable of diagnosing both types of epilepsy.

For animals with a probable diagnosis of idiopathic epilepsy (i.e. those animals that fulfil Tier 1 level of confidence for diagnosis - see Consensus Proposal: Diagnostic approach to epilepsy in dogs), many of the differential diagnoses associated with structural epilepsy, in particular large malformations and neoplastic causes, are relatively straightforward to identify [6, 17, 18]. However, several are associated with subtle changes that may be easily missed without adequate resolution scanning and careful interpretation. The most common of these are listed in Table 4. It must also be remembered that any lesion identified is not automatically epileptogenic in nature and other evidence (e.g. EEG, seizure history) may be required to demonstrate this [19].

Table 4 Differentials for idiopathic epilepsy that may require high resolution imaging to identify
Full size table

Identification of lesions which are the consequence of seizures

Longitudinal studies of epileptic humans suggest that 10 % of newly diagnosed patients and 25 % of those with chronic active epilepsy develop significant cerebral, hippocampal or cerebellar atrophy over 3.5 years [20]. More acute changes secondary to seizures have also been reported (Fig. 1) and it is important that imaging techniques are able to differentiate these resultant, reversible changes from those that may be the cause of seizures. Most commonly, changes that are the result of seizures are found as T2-weighted hyperintensities predominantly in the piriform and temporal lobes, as well as the cingulate gyrus and hippocampus [21]. These changes resemble those reported in humans and are likely to represent a mixture of cytotoxic oedema and gliosis [21]. In some cases mild contrast uptake may also be apparent [22]. In general these changes are diffuse, relatively extensive, and their characteristic location makes it straightforward to distinguish them from epileptogenic lesions with either high-field or low-field scanners. However sometimes in can be difficult to ascertain if the changes are cause or effect for example in VGKC-complex/LGI1 antibody-associated limbic encephalitis in cats (Fig. 2) [23]. Cerebrospinal fluid analysis can be unhelpful because post-ictal pleocytosis can occur [24]. In ideal circumstances it would be preferred to repeat imaging in the post ictal period and also assess changes in brain volume/ atrophy however available finances can limit this opportunity. In those patients with whom some doubt may remain, however, the most useful procedure for identifying post-ictal MRI changes is to repeat the scan at a later date, since these changes resolve usually within 16 weeks [21].

Fig. 1
figure 1

Post-ictal changes in the temporal and parietal lobe. Images obtained in a 1.5 T Siemens Symphony, Erlangen, Germany. Post-ictal oedema in the temporal lobe (short white arrow), hippocampus (long white arrow) and cingulate gyrus (yellow arrow) in a 2 year male English Bulldog that presented in status epilepticus

Full size image
Fig. 2
figure 2

Hippocampal changes in an 8 month male neutered Oriental Shorthair presented with status epilepticus. a Transverse TW2 at level of pituitary gland. There is hyperintensity of the right temporal lobe (red arrow) (b) Transverse FLAIR at level of pituitary gland also demonstrating hyperintensity of the right temporal lobe (red arrow) (c) Transverse TW1 at level of pituitary gland. There is slight gadolinium contrast enhancement in the mesial temporal lobe. Images reproduced with the kind permission of Dr Ane Uriarte . The cat was suspected to have limbic encephalitis

Full size image

Providing data for further research into the pathogenesis and treatment of seizures

In humans, much attention has focussed on the hippocampus because temporal lobe epilepsy (TLE) is the most common cause of complex focal epilepsy, and mesial temporal sclerosis (i.e. severe neuronal cell loss and gliosis in the medial portion of the temporal lobe and particularly in the hippocampus) is a major pathological finding, occurring in roughly 50 % of TLE patients [25]. The pathogenesis of mesial temporal sclerosis is multifactorial and includes genetic factors and molecular events such as channelopathies, activation of NMDA receptors, and other conditions related to Ca(2+) influx into neurons and imbalance of Ca(2+)-binding proteins [26]. There has been much debate as to whether these changes are the cause or the effect of seizures. Most significantly, the surgical removal of these regions in patients with an electroencephalographic (EEG) diagnosis that confirms their location as the source of seizure activity results in significant improvement in seizure control in up to 80 % of patients [27, 28]. The current diagnosis of hippocampal sclerosis in humans requires specific positioning of slices in order to define the hippocampus accurately, together with a considerable body of research defining the range of normal volumes in healthy individuals. These techniques for hippocampal measurement have been established for many years and TLE is one of the more common homogenous forms of epilepsy, so adequate numbers of patients are available for studies [20].

Whether hippocampal volume loss and mesial temporal sclerosis is a parameter that should be assessed in dog has yet to be established (Fig. 1). Hippocampal atrophy has been demonstrated in rodent models [29] and in familial spontaneous epileptic cats where EEG features suggested TLE [16]. Reduced volume of the hippocampus / hippocampal atrophy has been demonstrated in epileptic dogs [15]. Furthermore histopathological changes consistent with hippocampal sclerosis have been well described in epileptic cats [28, 30, 31] (Figs. 2, 3). For these reasons, as well as the recognition that hippocampal sclerosis represents a common surgical target in the treatment of human epilepsy, it appears prudent to evaluate the hippocampus accurately in animal patients with epilepsy. Therefore routine MR evaluation of the epileptic subject should at least include a visual assessment of the hippocampus for atrophy, asymmetry in size, loss of defined morphologic structure, increased T2W or T2W Fluid Attenuated Inversion Recovery (FLAIR) signal and decreased T1W signal [15, 32]. Hippocampal T2W hyperintensity is well correlated with pathology and hippocampal sclerosis and measurement of the T2 relaxation time (T2 relaxometry) can provide an objective measure in humans but has not been assessed in dogs or cats [32]. There is an argument that volumetric studies should be performed in veterinary patients (Table 5) and recent studies have defined the range in normal animals [33]. However making volumetric measurements is a labour intensive process requiring high resolution MRI and personnel training [33]. Currently this is only used as a research tool, although in the future automated atlas-based segmentation may make hippocampal volumetry more routine. Even in humans where hippocampal volumetry has established utility, the time demands and required technical skills mean that it has been difficult to integrate into clinical practice [34]. Consequently patients with a surgically resectable lesion may be missed. This has led to the development of automated software which will compare an individual patient’s regional brain volumes with a normative database, correcting for sex, head size, and age [34]. Establishing automated software in veterinary patients is challenging due to difficulties in automatic brain extraction algorithms arising from the great variation is head shape and brain size and conformation. Establishment of reference ranges for the three basic canine brain shapes (dolicocephalic, mesaticephalic and brachycephalic) may represent a suitable compromise. Before making a recommendation of measurement of hippocampal volumes in veterinary patients it should be remembered that hippocampal sclerosis is not applicable to all idiopathic generalised epilepsies in humans especially if the epileptogenic focus is not the temporal lobe [35]. Repeated seizures will affect other structures pathologically including the amygdala, cerebral neocortex and the cerebellum [20].

Fig. 3
figure 3

Hippocampal changes in a 22 month male neutered Oriental Shorthair with drug-resitant epilepsy. Images obtained in a 1.5 T MRI (Siemens Symphony, Erlangen, Germany) 12 months after the images in Fig. 2. Despite an initial course of corticosteroids and polypharmacy with multiple anti-convulsants the cat seizured on an almost daily basis. a Dorsal T2W orientated perpendicular to long axis of the hippocampus. b Transverse T2W orientated parallel to the long axis of the hippocampus. c Dorsal FLAIR orientated perpendicular to long axis of the hippocampus. d Transverse FLAIR orientated to long axis of the hippocampus. e Dorsal T1W 3D images 1 mm slice thickness orientated perpendicular to long axis of the hippocampus. f Dorsal T1W orientated perpendicular to long axis of the hippocampus post gadolinium. On FLAIR and T2W images there is reduction in volume and a hyperintensity of the hippocampus (yellow arrows). With the TIW 3D images it is possible to appreciate loss in definition between the white and grey matter in addition to reduction in volume of the hippocampus (blue arrow) There is no abnormal enhancement with gadolinium contrast

Full size image
Table 5 Reasons why it may be appropriate to perform volumetric studies on hippocampus or other potentially epileptogenic areas
Full size table

Existing MRI protocols

Current protocols vary substantially between institutions. Polling of members of the international veterinary epilepsy task force determined that all protocols currently include imaging in at least two orientations (transverse and sagittal) and the majority in three planes (dorsal, typically orientated parallel to hard palate rather than perpendicular to the long axis of the hippocampus). T2W, T2W FLAIR and T1W images pre and post paramagnetic contrast (gadolinium based) are included as standard in most protocols used by specialists who are active in the veterinary field. This differs from human epilepsy-specific MRI protocols where routine administration of gadolinium contrast is considered to provide little advantage for idiopathic or TLE and is reserved for patients in whom there is concern for tumour, vascular malformations, inflammation, and infectious disease or when these are suspected based on review of non-contrast studies [35]. Routine administration of gadolinium contrast in veterinary medicine has been questioned [36]. Other sequences currently included in “veterinary brain protocols” vary between institutions and may include Gradient Echo (GE), T1 weighted Inversion Recovery (T1WIR), Diffusion Weighted Imaging (DWI) and Short Tau Inversion Recovery (STIR) or other fat suppression techniques.

This variation between institutions suggests a need for a uniform veterinary epilepsy-specific MRI protocol that can provide a solid platform for clinical communication and comparability of case definition between research studies. There is also an argument for an MRI protocol that is optimized for epilepsy evaluation facilitating more detailed examination of areas susceptible to generating and perpetuating seizures such as the frontal and temporal lobes and other structures likely to be evaluated at post-mortem in patients who have died. Such a protocol must acknowledge financial constraints, be tailored for low or high field machines and also complement pathological studies.

Consensus on epilepsy-specific MRI protocol

There is a need for a standardized veterinary epilepsy-specific MRI protocol which will facilitate more detailed examination of areas susceptible to generating and perpetuating seizures, complement pathological studies, is economical, simple to perform and can be adapted for both low and high field machines. Standardisation of imaging will improve clinical communication and uniformity of case definition between research studies. We propose the following protocols (Tables 6 and 7). During protocol set-up, it is recommended that different parameters (such as flip angle) are trialed in order to obtain the optimal balance between grey-white matter contrast and SNR (for information on MR parameters for 0, 2, 1.5 and 3T see Additional files 1, 2 and 3). Both protocols start with obtaining a sagittal sequence. Due to the difference in anatomical definition this is a T2W sequence in high field machines and T1W sequence in low field machines. In addition to identifying gross structural pathology the sagittal images allow assessment of cerebellar atrophy according to the protocol described by Thames and others [37]. Using parasagittal images the long axis of the hippocampus is identified (Figs. 4, 5, 6, 7, 8 and 9). The hippocampus forms the medial wall of the temporal horn of the lateral ventricle and is delineated on parasagittal images by the contrasting cerebrospinal fluid. After identification of the hippocampus, T2W and sequences are orientated parallel and perpendicular to the long axis of the hippocampus (Figs. 4 and 7). T2W and FLAIR are acknowledged to be optimal for detection of epileptic lesions in humans in particular hippocampal changes (Figs. 2 and 3) and therefore in humans two FLAIR sequences would be obtained [38], however, it is recognized that performing two FLAIR sequences may increase scanning time significantly therefore we recommend that at a minimum a dorsal FLAIR sequence perpendicular to the long axis of the hippocampus is obtained with an option for an additional transverse sequence parallel to the long axis of the hippocampus. In high field scanners a transverse gradient echo or similar sequences sensitive to detection of hemosiderin and / or calcification should be obtained. Like the other images this transverse image is also orientated parallel to the hippocampus. In low field scanners additional T1W sequences are recommended (Table 5). Some high field machines may be able to obtain good resolution 3D TW1 images (Figs. 3, 8 and 10). For these the acquired slice thickness is 1 mm or less giving improved chance of lesion detection, better white and grey matter definition and can be processed after imaging into any anatomical plane including oblique. Furthermore this will facilitate volumetric measurements and to enable co-registration with other modalities or sequential MRI examinations [13, 38]. If this is not possible then a dorsal T1W sequence oriented along the long axis of the hippocampus is suggested. As indicated above there is an argument against routine paramagnetic contrast administration however it is acknowledged that many veterinary neurologists would feel a MRI study of an epileptic patient was incomplete without this therefore these sequences are an optional extra. However if pathology was detected in the unenhanced study, post-gadolinium sequences would be indicated (Fig. 10). Recommended slice thickness is 3 mm or less for high field machines and 4 mm or less for low field machines. Such a protocol would give 6–7 sequences for a high field machine and 6–7 sequences on a low field machine (not including optional paramagnetic contrast enhancement).

Table 6 Proposed epilepsy-specific MRI protocol for a high field machine
Full size table
Table 7 Proposed epilepsy specific MRI protocol for a low field machine
Full size table
Fig. 4
figure 4

Parasaggital slice in a veterinary epilepsy-specific protocol for 1.5 T MRI scanner. T2W parasagittal image of the brain demonstrating a planned sequence parallel (yellow dotted line) and perpendicular (red solid line) to the long axis of the hippocampus. Images obtained in a 1.5 T MRI (Siemens Symphony, Erlangen, Germany)

Full size image
Fig. 5
figure 5

Veterinary epilepsy-specific protocol for high field MRI. Images obtained in a 1.5 T MRI (Siemens Symphony, Erlangen, Germany). Triplet of MR images illustrating the positon of the parasagittal slice containing the hippocampus. Left. T2W parasagittal section demonstrating the hippocampus for sequences orientated relative to the long axis. Middle. Dorsal FLAIR of the brain at the level of the orbits illustrating the position of the parasagittal slice (green line). Right T2W transverse of the brain at the level of the hippocampus illustrating the position of the parasagittal slice (green line)

Full size image
Fig. 6
figure 6

Variation in appearance of the hippocampus in different skull shapes. a brachycephalic vs (b) mesocephalic vs (c) dolicocephalic with orientation of transverse scans parallel to the long axis of the hippocampus

Full size image
Fig. 7
figure 7

Veterinary epilepsy-specific protocol for low field MRI. T1W parasagittal image (left) of the brain demonstrating a planned sequence orientated parallel (green line) and perpendicular (red solid line) to the long axis of the hippocampus. It is easier to identify the hippocampus in T1W images from a low field machine. For comparison the corresponding T2W parasagittal images are included (right). Images obtained in 0.2 T MRI (Esaote Grande, Genova, Italy)

Full size image
Fig. 8
figure 8

Veterinary epilepsy-specific protocol for high field MRI. The imaging time for 6 sequences (Figs. 8 and 9) on a 1.5 T MRI was 45 min. The subject was an epileptic 16 month female Cocker spaniel (a) and (b) Transverse and parasagittal T2W image illustrating slice orientation. c Dorsal T2W orientated perpendicular to long axis of the hippocampus (d) Dorsal FLAIR orientated perpendicular to long axis of the hippocampus (e) Dorsal T1W 3D images1mm slice thickness orientated perpendicular to long axis of the hippocampus

Full size image
Fig. 9
figure 9

Veterinary epilepsy-specific protocol for high field MRI. a parasagittal T2W image illustrating slice orientation. b Transverse T2W orientated parallel to the long axis of the hippocampus. c Transverse FLAIR orientated parallel to the long axis of the hippocampus. Images obtained in a 1.5 T MRI (Siemens Symphony, Erlangen, Germany)

Full size image
Fig. 10
figure 10

Representative MRI from a 2.95 kg 5 year female entire Chihuahua dog that underwent a diagnostic investigation for cluster seizures. a Parasagittal image demonstrating the hippocampus and the planned imaging perpendicular to the long axis (b) Dorsal FLAIR images orientated perpendicular to long axis of the hippocampus demonstrating hyperintensity in the frontal lobe (short arrow). Although this protocol is optimised for detection of hippocampal lesions visualisation of other pathology is not compromised. c Dorsal T1W 3D images 1 mm slice thickness orientated perpendicular to long axis of the hippocampus. The scrolled structure of the hippocampus is clearly defined despite the small patient size. Furthermore the demarcation between white and grey matter can be appreciated (long arrow). d Post gadolinium T1W images are obtained in further investigation of the frontal lobe pathology. The patient was diagnosed subsequently with necrotising encephalitis. Images obtained in a 1.5 T MRI (Siemens Symphony, Erlangen, Germany)

Full size image