Abstract
Objectives
The aim of this study was to evaluate the MRI findings in cats following traumatic brain injury (TBI), and establish which features identified might correlate with prognosis.
Methods
The magnetic resonance images of 30 cats with clinical signs attributed to the brain following TBI were retrospectively reviewed to assess the imaging features seen and their correlation with outcome.
Results
Twenty-one of 30 cats had a good outcome (full recovery or minor ongoing neurological deficits); the other nine either died or had ongoing neurological deficits that significantly affected quality of life. There was evidence of parenchymal injury in only 20/30 cats, including 8/9 that had a poor outcome. The frequency of bilateral or multifocal parenchymal lesions on T2-weighted imaging, and mass effect, particularly caudal transtentorial herniation, was statistically significantly higher in patients with a poorer outcome. Extra-axial haemorrhage was not identified. Concurrent soft tissue injury was noted in all cats, with a poorer prognosis statistically associated with a peripharyngeal pattern of injury and orbital trauma.
Conclusions and relevance
The number of cases with MRI-identified brain pathology, the patterns of injuries identified and the specific imaging features that appear to have effects on prognosis differ in this series of cats from those published in similar case series of dogs. Further work is needed to establish if cat-specific guidelines are needed to assess brain MRI following trauma.
Introduction
Traumatic brain injury (TBI) is recognised in feline patients following trauma, such as road traffic accidents, falling from a height and punctures due to bite wounds. 1 Advanced imaging, for example MRI, is frequently performed in small animal patients following TBI, but there is a paucity of studies considering the utility of this technique in cats and describing the most relevant features to assess.
The use of MRI in dogs following brain trauma has been assessed in the literature recently, with two studies noting that MRI features had an association with outcome. The first series of 50 dogs demonstrated that a greater area of affected brain parenchyma, 2 a greater degree of midline shift and a lower MRI score (using a scoring system modified from a human classification) 3 was correlated with poorer clinical outcome follow-up scores and that the modified Glasgow coma score is lower in dogs with evidence of brain herniation. Also in this study, development of post-traumatic epilepsy in dogs was associated with a greater percentage area of brain parenchyma affected on the magnetic resonance (MR) images. A second series of 18 dogs showed that dogs with parenchymal injury affecting the caudal cranial fossa had poorer outcomes (death, euthanasia or severe persistent neurological deficits). 4 The pattern of brain injury seen in dogs was proposed to differ from that seen in humans, 5 with the most common distribution for extra-axial haemorrhage being a coup-site subdural haemorrhage; the low number of cases with epidural haemorrhage had a poorer outcome, necessitating surgical decompression or euthanasia.
The pattern of injury in cats may be different to that in dogs as the inciting trauma may differ and the size and anatomical conformation of the head (particularly temporal musculature and skull) differs, so it is unlikely that prognostic factors identified in dogs will be the same in cats. It is known that cats are at a lower risk of developing post-traumatic epilepsy than dogs,6,7 supporting the theory that cats have a different outcome to that seen in dogs.
The aims of this study were to: collate a case series of cats with evidence of neurological deficits that have undergone brain MRI following head trauma; review the MR images with a view to identifying specific MR features that are seen with trauma on both low- and high-field magnets; and assess whether any of these features are associated with the outcome.
Our hypotheses were that the pattern of MRI injuries in cats will differ from those described in dogs and the MRI features with prognostic significance in dogs will not be predictive of the outcome in cats.
Materials and methods
This is a retrospective descriptive study, using case populations obtained from two institutions. The medical records of both institutions were reviewed from 2000–2016 for all cats with suspected or known TBI that underwent MRI investigations of the brain. The inclusion criteria for this study were: (1) history compatible with TBI; (2) neurological examination with evidence of central neurological dysfunction localising to the brain; (3) no pre-existing neurological disease; and (4) MRI performed within 1 week of injury.
For each case, the clinical data recorded included signalment, suspected/known cause of trauma, and general physical and neurological examination findings at admission.
The follow-up, including repeat clinical and neurological examination at institution or by the referring veterinary surgeon, was recorded if available. A follow-up telephone conversation was held with the owners if no re-examination had been performed. Patient follow-up was categorised into four groups: (1) complete recovery; (2) ongoing mild neurological deficits that did not affect the quality of life (as perceived by the owner); (3) ongoing major neurological deficits that were perceived to affect the quality of life of the patient; and (4) death (either spontaneous or euthanased as a result of the brain injuries sustained in this trauma). For the purpose of analysis, grades 1 and 2 were considered to be a good outcome, and 3 and 4 were considered a bad outcome.
MR scans were obtained at the respective institutions with a 1.5 Tesla (T) closed-bore magnet (GE Signa Echospeed; GE Medical Systems) and a 0.4 T open magnet (Hitachi Aperto; Hitachi Medical). Sequences varied between patients, and are listed alongside the specific imaging parameters in a table in the supplementary material.
Gadolinium-based contrast medium was administered at 0.1 mmol/kg body weight (variably gadoterate meglumine [Dotarem; Guerbert], gadobenate dimeglumine [Multihance; Bracco] and gabutrol [Gadovist; Bayer]), with T1-weighted (T1W) images repeated following contrast administration in various planes.
The MR images were retrospectively reviewed collaboratively by two board-certified radiologists (AC and RD) blinded to the clinical presentation and outcome.
The features assessed included the following:
(1) Presence or absence of parenchymal pathology (contusion, oedema) indicated by regions of T2-weighted (T2W) hyperintensity. Location was defined as unilateral vs bilateral, focal vs multifocal and rostral/middle vs caudal cranial fossa (ie, brain parenchyma affected is located rostral to a line drawn between the rostral tip of the tentorium cerebelli and the caudal bony margin of the infundibular recess, or caudal to it).
(2) Percentage of affected brain tissue: subjectively estimated affected percentage of the brain measured on the transverse T2W images compared with the overall volume of brain tissue in categories 0%, 0–1%, 1–5%, 5–10%, 10–20% and >20%.
(3) Presence or absence of parenchymal haemorrhage, based on presence of gradient echo (GE) T2* susceptibility artefact and/or T1W hyperintensity in subacute cases.
(4) Presence or absence of parenchymal contrast enhancement.
(5) Presence or absence of any mass effect, determined by herniation (subdivided into transtentorial, transforaminal and transcalvarial) or midline shift. If midline shift was present, the displacement (in mm) of the current location of the midline of the brain from a line taken from the falx to the midline of the skull base on a transverse image was recorded (Figure 1).
(6) Presence or absence of suspected intracranial extra-axial haemorrhage (indicated by thickening of the T2W and/or fluid-attenuated inversion recovery hyperintensity surrounding the cerebral cortex without enhancement).
(7) Presence or absence of potential cranial nerve trauma, indicated by evidence of cranial nerve swelling, nerve T2W hyperintensity or extra-axial pathology in this region.
(8) Presence or absence of cranial vault fractures, categorised as: 0 (no fracture); 1 (hairline or insignificantly displaced fracture); and 2 (significantly displaced fracture).
(9) Additional injuries were noted: frontal sinus injury, nasal injury, orbital injury, tympanic bulla injury, location of soft tissue injury. (10) For each variable, differences in the frequency of observations in cats with a poor outcome (grade 3–4) and in cats with a good outcome (grade 1 or 2) were tested for significance (P <0.05) using Fisher’s exact test.
Procedure for measuring midline shift. The measurement in millimetres is made of the red horizontal line, which is the distance between the maximum deviation of the midline structures from the blue line (expected midline, a line placed from falx to centre of the pituitary fossa)
Results
Thirty cats met the inclusion criteria, 14 from the institution with a low-field magnet and 16 from the institution with a high-field magnet.
Clinical data
There were 30 cats in the study population: 20 males (two entire, 18 neutered) and 10 females (all neutered). Breed distribution was 23 domestic shorthairs, two domestic longhairs and one each of Russian Blue, Savannah, Oriental Shorthair, British Shorthair and Siamese breeds. The ages ranged from 6 months to 9 years (median 2 years, interquartile range [IQR] 1–4.5 years; for all of the following measurements, median and upper and lower quartiles are given, unless stated otherwise). Four cats had a dog fight as the cause of injury; the remaining 26 were road traffic accidents (note that in four cases the injuries were not witnessed and the cause was speculated).
All but two cats were managed medically with supportive treatment, including fluid therapy, antibiotics, pain relief (opiates/non-steroidal anti-inflammatory drugs) and feeding support. Mannitol was administered in 10/30 patients. Two patients underwent decompressive rostrotentorial craniectomy. Patients were discharged 1–30 days after admission (median 5.5 days; IQR 3–10 days).
Regarding outcome, follow-up was maintained until complete recovery, death or the last recorded veterinary visit if this was several years since discharge. Follow-up time ranged from 0 days to 11 years; median 6 months (IQR 2–17.25 months). Excluding those patients that died from their injuries, the follow-up period was between 1 month (if normal at this time) and 11 years, with a median of 7.5 months (IQR 3–33 months). Of the 24 surviving patients, 14 had a clinical and neurological re-evaluation within the first 1–3 months. The remaining 10 patients had a telephone follow-up during this period, nine of which had complete recovery of clinical signs before or shortly after discharge. The final patient was not re-examined, however, had no improvement in clinical signs described by the owner on telephone follow-up. Of those surviving cases with a longer follow-up beyond 3 months (19/24 cases), none changed category from good (1 or 2) to bad (3 and 4) or vice versa for the duration of their follow-up. Duration of hospital stay did not correlate with outcome (mean 8.2 days for those with a good outcome compared with a mean of 9 days for the five cases that survived to discharge but that had a poor outcome).
When categorising outcome, 21/30 (70%) had a good outcome: 17 were grade 1 (return to normal) and four were grade 2 (minor ongoing deficits not affecting quality of life). The remaining nine (30%) cats had a poor outcome: three were grade 3 and survived but had ongoing deficits that affected their quality of life, and six were grade 4 (death). Of the patients that died, four were euthanased: two patients in the peritrauma period due to remaining in a comatose state; and two a period after discharge (one at 5 months for progressive deterioration of neurological signs, persistent blindness and development of central diabetes insipidus, and one at 6 months for status epilepticus). The remaining two deaths occurred spontaneously in the peritrauma period due to respiratory arrest: for one we presumed a neurological cause of arrest; however, the other had evidence of concurrent pulmonary contusion and therefore it is unclear if the arrest was of neurological or pulmonary origin. Only one of the patients was euthanased immediately after MRI owing to severe changes on neurological presentation and MRI; all other cases in the bad (grade 3 or 4) category had spontaneous death or euthanasia after an appropriate duration to allow spontaneous recovery; therefore, owner desire to treat was considered to have minimal effect on the outcomes presented.
The remaining concurrent injuries that necessitated treatment are detailed in Table 1 but did not affect overall outcome.
Type and number of concurrent injuries identified in this population of cats
There was a good outcome in 9/14 cases imaged with a low-field system, and 12/16 of those imaged with a high-field system.
MRI examination
MRI was performed 12 h to 7 days following the time of trauma (mean 1.9 days). The sequences acquired in each case are listed in the supplementary material.
Parenchymal injury
Twenty of 30 (67%) cats had evidence of parenchymal T2W hyperintensity, including 8/9 cases (89%) with a poor outcome and 12/21 (57%) with a good outcome. No parenchymal change was seen in the remaining 10 cases (33%): 5/14 cases were examined with a low field and 5/16 cases were examined with a high field (findings are summarised in Table 2).
Outcome in cases demonstrating specific MRI features of the parenchyma
Significant (P <0.05)
T2W = T2-weighted; GE = gradient echo
Parenchymal change was multifocal in 10/30 (33%) cases, of which four had a good outcome and six a poor outcome (five were grade 4: died). Bilateral distribution was also seen in 10/30 (33%); all but one of the multifocal cases were bilateral – again four of these had a good outcome. Presence of both multifocal and bilaterally distributed lesions achieved statistical significance.
When present, parenchymal pathology was noted in the caudal cranial fossa (caudal to the osseous tentorium cerebelli) in three cases (cerebellum only in two cases, and one in both pons and cerebellum). No case had changes in the caudal cranial fossa only, it was always multifocal and in combination with a lesion rostral to the osseous tentorium cerebelli. Of the three cases with caudal cranial fossa pathology, there was one with a poor outcome (which had cerebellum and pons location); this did not achieve statistical significance.
Figure 2 illustrates the effect of estimated percentage area against outcome.
Bar chart illustrating the distribution of estimated percentage of affected area of the parenchyma between good and poor outcomes. (a) All cats; (b) same data presented but having removed any cats in which prior administration of mannitol could have had an effect on the area of brain parenchyma affected
All but one case had a GE T2* sequence performed, of which 13 had signal void indicating the presence of haemorrhage: six had a poor outcome, which did not achieve statistical significance. Six cases had spin echo T2 parenchymal pathology identified; however, a signal void indicative of haemorrhage was not identified on T2*. The high-field system and low-field system each ‘missed’ haemorrhage in three cases (Figure 3).
Transverse T2*-weighted images of the forebrain from two cats. Examples of parenchymal injury with haemorrhage in (a) high field (a 2.5-year-old male neutered domestic shorthair cat imaged 2 days after a road traffic accident, outcome grade 2) and (b) low field (a 2-year-old male neutered domestic shorthair cat imaged within 24 h of a road traffic accident, outcome grade 4) demonstrating that T2* effects may appear slightly differently with a low-field scanner, with a more layered appearance than a uniform signal void (white arrows point to the signal void)
A degree of mass effect, of any type, was noted in eight cases – five of which had a poor outcome. Transtentorial herniation was always seen in a caudal direction and was present in four cases, all with a poor outcome; both the presence of some mass effect and specifically caudal transtentorial herniation achieved a statistical significance for association with a bad outcome. Seven of the eight cases with mass effect had midline shift that ranged from 0.5–2.6 mm (mean 1.3 mm). Size of shift did not correlate with outcome: the smallest shift (0.5 mm) was seen by a cat in the good outcome group and a cat in the poor outcome group; similarly, the largest size of shift (2.6 mm) was seen by one cat in each group.
Contrast (performed in 25 cases) uptake was noted in the parenchyma in only five cases (three with a good outcome, two with a poor outcome), all of which also had T2W pathology identified, but there was subjectively no correlation with extent of the T2W abnormality. For the five cases showing abnormal contrast uptake the MR scan was performed 1–4 days after the suspected time of trauma (median 3 days), slightly more than the median time to MRI of cases without uptake of 1.25 days (range 0.5–7 days).
There was no convincing evidence of extra-axial haemorrhage in any patient.
Extra-parenchymal head injury
MRI identified cranial vault fractures (Figure 4) in 12 cases, of which 10 were grade 1 (insignificant displacement) and only two were grade 2 (significantly displaced). Seven of the 12 cases with cranial fractures had a good outcome; of the two cats with a significantly displaced fracture (both requiring surgery), one had a good outcome (grade 2, blind in one eye) and the other a poor outcome (grade 3, bilateral blindness and persistent circling). Five cases had pathology affecting the neural tissues adjacent to the skull base, with suspected or potential trauma that could affect the cranial nerves or their cell bodies in this region (optic chiasm in two, pons in two, diffusely along the floor of the cranium in two and cavernous sinus in one). Interestingly, four of five cases with cranial base trauma ± fractures had a poor outcome. Optic nerve/optic chiasmal (Figure 5) or orbital trauma suspected to affect the optic nerve was seen in seven cases, all of which were blind. Orbital trauma was seen overall in 11 cases, seven of which had a poor outcome, with orbital trauma statistically significantly associated with a poor outcome. Findings are summarised in Table 3.
Examples of fractures. T1-weighted transverse images of the cranium of two cats indicating different fracture grades. (a) Grade 1 fracture in a 2-year-old male neutered domestic shorthair cat imaged <1 day after a road traffic accident (outcome grade 4) with a hairline fracture of the cranium with minimal displacement (white arrow) and (b) grade 2 fracture in a 4-year-old male domestic longhair cat (outcome grade 3) imaged 2 days following a dog attack with a fragment of bone (short arrows) displaced axially into the brain parenchyma from the cranium (long arrows)
T2-weighted transverse images of a 9-month-old male entire Russian Blue cat imaged 1 day after a road traffic accident (outcome grade 4), taken at the level of (a) the orbital fissure demonstrating the asymmetry of the skull base (arrows) and (b) the optic chiasm (arrows), showing an example of an optic chiasmal injury, with a swollen hyperintense optic chiasm, adjacent to traumatic skull-base asymmetry
Outcome in cases demonstrating specific non-neurological injuries of the head
Significant (P <0.05)
Soft tissue trauma was always identified, with radiologists noting that post-contrast T1W sequences were more sensitive for the detection of muscle pathology than T2W sequences were. Soft tissue trauma most commonly affected the temporal muscles. Fourteen of 30 cats demonstrated specific patterns of trauma (either in addition to temporal muscle injury or in isolation) affecting the either the rostral/nasal structures with a shearing-type injury (eight cases; Figure 6a,b) and/or the peripharyngeal tissues with swelling/haematoma (10 cases; Figure 6c). Identification of the peripharyngeal pattern was statistically associated with a bad outcome; however, the presence of bilateral soft tissue injury was not.
Examples of the two distinct patterns of extraparenchymal injury. (a) (at the level of the rostral nasal cavity) and (b) (level of the caudal nasal cavity) are T1-weighted images of a 3-year-old domestic shorthair cat imaged 2 days after a road traffic accident (outcome grade 3) demonstrating the rostral pattern of injury, with nasal fractures that propagate into skull-base fractures (with arrows indicating the fracture line). (c) is a T2-weighted (T2W) transverse image at the level of the pharynx of a 5-year-old male neutered domestic shorthair cat imaged 1 day after a road traffic accident (outcome grade 3) demonstrating a peripharyngeal pattern of soft tissue injury, with T2W soft tissue hyperintensity indicated by arrows
Specific locations of extracranial injury are noted in Table 3.
Discussion
There was a greater proportion of cats in this study with immediate onset of intracranially located neurological signs following trauma that had no visible pathology on MRI than has been reported in dogs.2,4 It is not known whether this is a species difference or is related to the smaller size of the cat that challenges MR image resolution. The extent of the area affected appears only weakly associated with outcome – although there were more good outcomes with the lowest percentages of area affected (0–1%) and poor outcomes with the two cases with the highest percentage area affected (>20%) there was a good outcome for the only two cases with the second highest grade (10–20%), which differs from dogs, 2 where increasing percentage area affected was associated with a poorer outcome. Although multifocal and bilateral lesions (note that the majority of multifocal lesions were also bilateral) were statistically associated with a bad outcome 8 , there was no significance of caudal cranial fossa involvement, which also differs from the findings in dogs 4 and humans, 9 although the number of cases with caudal cranial fossa involvement in this study was very small.
GE T2* signal void indicative of haemorrhage was more likely to be seen in our cases with a poor outcome; however, this did not achieve statistical significance. In people, the use of susceptibility-weighted imaging 10 to identify blood products has been associated with acute TBI severity and short-term outcome but has not consistently been associated with long-term outcomes. 11 There was a slightly greater proportion of cases in which haemorrhage was not identified with the low-field magnet (n = 3/13 [23%] vs n = 3/16 [19%]) for the high-field magnet, unsurprising given the inherent increase in sensitivity to susceptibility artefacts of a high-field magnet. When haemorrhage was identified with the low-field magnet it was subjectively more subtle than when seen with the high-field magnet. If identification of haemorrhage does prove to be a useful prognostic factor with future validation within a larger case cohort, then it is prudent to pay close attention to very subtle foci of signal void on GE T2*, particularly if using a low-field scanner.
Unfortunately, it is not possible to comment at this stage on the utility of contrast administration in assessing prognosis. Contrast medium was administered in 25/30 cases − 4/5 cases that did not have contrast administered were in the group with a poor outcome, which hampers assessment of whether presence or absence of contrast uptake has a prognostic effect. All the cases that did not have contrast medium administered had abnormality identifiable on T2W images, which suggests the radiologists at the time of image acquisition elected not to administer contrast because they had already identified changes they felt sufficient to make the diagnosis in that particular case and/or the patient condition precluded prolongation of the image acquisition time. Established human literature does not support the use of contrast MRI in the early days following TBI as it does not improve the number of lesions identified. 12 The cases with contrast uptake identified in the parenchyma had a slightly longer median time to MRI after the traumatic incident, which correlates with one study that only identified parenchymal contrast enhancement of traumatic lesions in human patients imaged more than 4 days following the injury. 13
Clear evidence of extra-axial haemorrhage was not noted in any patient, which is interesting considering that it is seen in humans 5 and dogs 4 relatively frequently after trauma.
Although a mass effect was noted with a greater proportion of cases with a poor outcome, there is no indication that size of midline shift can be used as a prognostic indicator as it might be with humans 5 or dogs 4 ; although similar to dogs, type of mass effect may play a role as caudal transtentorial herniation was only seen in cases with a poor outcome.
Severe bilateral soft tissue injury was more commonly seen in cases with a poor outcome (78% of cases with a poor outcome vs 42% of cases with a good outcome, although this did not achieve statistical significance), probably reflecting the severity of trauma in these cases. 14 The small size of a cat’s head compared with most dogs means that blunt trauma is likely to impact on many structures simultaneously. Two patterns of soft tissue injury were seen: firstly, rostral with nasal/orbital/frontal sinus injury combined with brain injury with some of these nasal injuries propagating to skull-base fractures with associated skull-base parenchymal and potential cranial nerve injury. A second, more caudally located pattern of injury was trauma in the peripharyngeal region surrounding the pharynx and retropharyngeal lymph nodes, which, interestingly, occurred significantly more often in animals suffering a poor outcome, although a cause for this is unclear. Recognition of these patterns helps radiologists to identify all clinically relevant concurrent injuries readily. Both patterns of injury may be seen in the same patient, most likely representing the initial blunt impact and a second opposing injury caused by impacting on a second surface.
The limitations of this study are largely due its retrospective nature and to the fact that two institutions with different types of MR scanner were involved. The MRI protocol varied between cases and case data were not always complete. Early follow-up was always achieved; however, although those cases with ongoing injury always had several months of follow-up documented, the possibility of a very-late-stage complication could not be excluded. Additionally, different clinicians may manage cases differently, leading to differences in the time taken to recommend advanced imaging for the patient and potentially differing clinical outcomes. In particular, treatment of the case with mannitol prior to imaging was not standardised – eight cases had mannitol administered within a window that might have had clinical impact prior to MRI; however, it is not clear from the literature if treatment with mannitol has any effect on standard MR images.
Despite utilising the data from two institutions with busy neurology and trauma caseloads, the total case number included remains relatively small, and trends identified require further investigation with greater numbers – with such a small group of affected cats, even results that have achieved statistical significance should be interpreted with caution. Comparison between this study and other studies in different species, while interesting, is to be interpreted with caution, given that the retrospective nature has led to differing inclusion criteria for this study.
Conclusions
Following TBI, cats appear to show a different pattern of injuries to dogs, and those factors identified as having potential for prognostic significance differ from dogs. There appears to be a poorer prognosis with the presence of bilateral and/or multifocal parenchymal changes; however, a caudal cranial fossa location of injury (noted to have poorer prognosis in dogs)2,4 was not statistically significantly associated with a poorer outcome. The percentage area of brain parenchyma may weakly associate with prognosis.
Mass effect was uncommonly seen and generally small, with the presence of caudal transtentorial herniation statistically associated with a poorer prognosis. As opposed to previously described dog populations, 4 extra-axial haemorrhage was not identified.
Supplemental Material
Supplementary_data_table – Supplemental material for Magnetic resonance imaging in 30 cats with traumatic brain injury
Supplemental material, Supplementary_data_table for Magnetic resonance imaging in 30 cats with traumatic brain injury by Abby Caine, Robert Brash, Luisa De Risio, Jan Van Dijk, Giunio Bruto Cherubini and Ruth Dennis in Journal of Feline Medicine and Surgery
Footnotes
Supplementary material
The following file is available online: Pulse sequence parameters.
Conflict of interest
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
