Correlating magnetic resonance findings with neuropathology and clinical signs in dogs and cats

Abstract

The histologic characteristics that are the basis for diagnosis of central nervous system conditions cannot be visualized directly using magnetic resonance (MR) methods, but clinical diagnosis may be based on the frequency and pattern of MR imaging signs, which represent predominantly the gross morphologic features of lesions. Additional quantitative MR measures of myelination, cell swelling, gliosis, and neuronal loss may also be used for more specific characterization of lesions. These measures include magnetization transfer ratio, apparent diffusion coefficient, and the concentrations or ratios of metabolites identified by spectroscopy. Confidence that an MR abnormality is responsible for the clinical signs depends primarily on the degree of correspondence between the site of the lesion and the neuroanatomical localization.

In general, magnetic resonance (MR) imaging is a sensitive, but non-specific method for the diagnosis of neural lesions.^1,2 Sensitivity is higher for masses, malformations, and inflammation while lower for chronic progressive neuronal degeneration. Differential diagnoses are usually generated based on a combination of imaging signs and patient data such as signalment, progression of clinical signs, and clinicopathological results. Similarly, confidence that an MR abnormality is responsible for the clinical signs depends primarily on the degree of correspondence between the site of the lesion and the neuroanatomical localization; however, for many common lesions, such as hydrocephalus, quadrigeminal cyst, caudal occipital malformation and brain atrophy, there is no consensus about how to correlate severity in MR images and the severity of clinical signs.

Towards more specific diagnosis of brain disease

The histologic characteristics that are the basis for definitive diagnosis of central nervous system conditions cannot be visualized directly using MR imaging, but clinical diagnosis may be based on the frequency and pattern of MR imaging signs, which for the most part represent gross morphologic features of lesions. For example, the presence of a dural tail and signal void suggestive of calcification in an extra-axial mass together with hyperostosis of the overlying bone can help distinguish a meningioma, in which these signs are common, from a peripheral nerve sheath tumor or a round cell neoplasm, in which this combination of signs would be unusual. Various reports have described the MR signs in patients with age-related degeneration,³ caudal occipital malformation syndrome,^4-6 arachnoid cysts,^7-9 hydrocephalus,^10-12 cerebellar degeneration, ^13,14 necrotizing encephalitis¹⁵, granulomatous meningoencephalitis,¹⁶ infarcts,^{17, 18} pituitary tumors,^{19, 20} and brain neoplasms.^21-27 These studies help estimate the prevalence of MR signs associated with these diseases, thus aiding interpretion of images of other patients. It may be that retrospective evaluation of gross and microscopic findings obtained from large populations of specific diseases for the frequency of potentially MR-visible abnormalities such as cysts, hemorrhage, and/or mineralization would also provide very useful data regarding the incidence of these abnormalities in imaged patients.^30-32

Knowledge of the prevalence of gross and histologic abnormalities in specific diseases is useful when interpreting MR images. ^{28, 29} For example, necrotizing encephalitis is an inflammatory disease of the brain due to unknown etiology.^{33, 34} Multifocal to coalescing lesions may be found throughout the brain, but are more commonly confined to the prosencephalon. Histologically, these lesions appear as rarefied parenchyma with infiltration of lymphocytes, histiocytes, and plasma cells.^{33, 34} Figure 1 shows the brain of a 1-year-old Maltese dog. Multifocal lesions are limited to the cerebrum and thalamus and are characterized by the presence of moderately contrast-enhancing parenchyma and meninges, as well as by the loss of parenchyma and the replacement by fluid. A meningoencephalitis is suspected from this pattern of multifocal enhancement. Diagnosis of necrotizing encephalitis is supported by the parenchymal loss, sparing of the caudal fossa, and signalment of the patient. In comparison, granulomatous meningoencephalomyelitis frequently causes lesions throughout the CNS but large regions of parenchymal loss are rare (Figure 2).³⁵

Figure 1.

Images of the brain of a 1 year old Maltese dog with necrotizing encephalitis. Extensive signal abnormalities are present throughout the gray and white matter of the left cerebrum and thalamus. Mild to moderate gadolinium enhancement is present in the cerebrum and overlying meninges. Focal parenchymal loss and replacement with fluid are visible in the left thalamus and temporal lobe. (A - T1-weighted image; B – T1-weighted image + gadolinium; C – T2-weighted image, D – gradient echo image)

Figure 2.

Images of the brain of a 6 year old dachshund with granulomatous meningoencephalomyelitis. Extensive signal abnormalities are present throughout the gray and white matter of the cerebellum on T2-weighted image. Strong ring enhancing lesions are present in the cerebellum. Swelling is suggested by the lack of visible cerebellar folia. (A - T1-weighted image + gadolinium; B – T2-weighted image)

Globoid cell leukodystrophy is a genetic, degenerative disease of the white matter of immature dogs.³⁶ Histologically, lesions appear as demyelination and loss of oligodendrocytes most severely affecting the centrum semiovale, corona radiata, and corpus callosum.³⁷ Large numbers of swollen macrophages (“globoid cells”) also occur. T2-weighted images show bilaterally symmetrical hyperintensity of the white matter with the centrum semiovale and corona radiata most severely affected (Figure 3). Gadolinium-enhanced T1-weighted images show enhancement of these regions suggestive of inflammation. The pattern of white matter involvement helps to distinguish leukodystrophy from other white matter diseases of puppies, such as canine distemper virus infection, which usually has a less symmetrical, multifocal appearance.

Figure 3.

T2-weighted image of the brains of a 16-week-old Cairn terrier with globoid cell leukodystrophy showing the bilaterally symmetrical increase in signal intensity of the white matter throughout the brain.

Other lesions remain difficult to distinguish. For example, gliomas and infarcts are intra-axial lesions that affect similar regions of the brain in older dogs. Histologically, gliomas appear as a proliferation of neoplastic cells with mass effect and rare hemorrhage.²⁹ Infarcts are associated with cell swelling and occasionally hemorrhage, followed by cell death, invasion by inflammatory cells and eventual parenchymal loss.¹⁷ Although they are readily distinguished histologically, gliomas and infarcts may appear similar in MR images. In theory, the degree of enhancement and the presence or absence of mass effect should aid diagnosis because gliomas commonly show moderate to strong enhancement, perilesional edema and mass effect while infarcts show mild if any contrast enhancement and little mass effect.^{17, 26} However, a recent study in which three blinded observers were requested to classify 38 conventional T1- and T2-weighted images as either infarct or glioma, over 10% of the infarcts were incorrectly interpreted to be gliomas.³⁸ Distinguishing glioma from infarct is difficult without the benefit of additional specific imaging sequences (Figure 4).

Figure 4.

Images A and B are of the brain of an 8 year old mixed-breed dog with an oligodendroglioma within the right side of the thalamus. The lesion shows strong ring-enhancement and the presence of a central cavity with signal characteristics distinct from that of CSF. Images C and D are of the brain of a 14 year old mixed breed dog with an infarct within the right temporal lobe. The lesion shows moderate to strong ring-enhancement, regionally extensive edema, and evidence of hemorrhage. (A & C – T1-weighted image + gadolinium; B &D – T2-weighted image)

There is potential for more specific diagnosis by using additional MR techniques complementary to the standard sequences. Examples include improved characterization of white matter disease using magnetization transfer imaging (MTI);^{39, 40} detection of cell swelling using diffusion-weighted imaging (DWI);^41-43 and assessment of neuronal loss, gliosis, and membrane turnover using magnetic resonance spectroscopy (MRS).⁴⁴

Magnetization transfer imaging (MTI)

MTI generates contrast by imaging the effect on signal intensity of the exchange of magnetization between protons in structural macromolecules and protons in water. By selectively saturating the signal from macromolecules, such as myelin, MTI can substantially increase tissue contrast and can improve sensitivity for the detection of disease.^{45, 46} MTI has been used to increase the sensitivity of lesion detection by increasing contrast for magneztic resonance angiography, for gadolinium-enhanced images, and for T2-weighted imaging to detect demyelinating disease.^{47, 48} The magnetization transfer ratio (MTR) is the quantitative expression of MTI which aides in both the detection and the characterization of white matter abnormalities. Use of the MTR is exemplified in imaging studies of multiple sclerosis (MS).^{39, 40} T2-weighted imaging alone of brain lesions in MS is insufficient to characterize the heterogeneity of plaques that may contain varying degrees of inflammation, demyelination, remyelination, axonal damage and gliosis.⁴⁹ In contrast, regional MTR analysis shows differences between plaques that reflect differences in their histology^{49, 51} MTR has also been used to differentiate demyelination from edema,³⁹ to measure myelin maturation,^{40, 53} to characterize periventricular hyperintense white matter disease in elderly patients,⁵⁴ and to quantify white matter lesions in trauma⁵⁵ and neurodegeneration.^{52, 56, 57}

MTR has been used to examine Wallerian degeneration in the visual system of cats.⁵² MTR values in white matter of the visual system increase as a result of Wallerian degeneration within two weeks after injury, which is earlier than abnormalities could be detected using either spin-echo imaging or light microscopy.

MTR has also been used to evaluate the maturation of brain myelin, the distribution and severity of abnormal myelination, and the effect of gene therapy on resolving myelination abnormalities in cats affected with the lysosomal storage disease alpha-mannosidosis.^{58, 59} In normal cats, regional increases in MTR of cerebral white matter are observed as myelin maturation occurs between 8 and 16 weeks of age. In contrast, cats with the dysmyelinating disease alpha-mannosidosis showed no such increase in MTR with time.^{58, 59} MTR has also been used in dogs to study distribution of abnormal myelination, using both region of interest analysis (Figure 5) and contour mapping.⁶⁰

Figure 5.

MTR of white matter regions (A) in a dog with globoid cell leukodystrophy (GLD) compared to a normal age matched dog using previously described methods.⁶⁰ Decreases in MTR were identified in the affected dog in all white matter tracts examined with the most significant differences occurring in the centrum semiovale, corpus callosum, and internal capsule (CWM, cerebellar white matter; SCP, superior cerebellar pednucle; OR, optic radiations; GL, corona radiata of gyrus lateralis; IC, internal capsule, CC, corpus callosum; CS, centrum semiovale). Changes observed on MRI are consistent with the microscopic changes seen in the white matter of the dog brain stained with Luxol fast blue (B; 25×).

The most frequent uses of MTI in clinical veterinary patients are probably following intravenous contrast administration to increase the conspicuity of enhancement of lesions or aid in performing time-of-flight angiographic studies. When using MTI in combination with gadolinium-containing contrast media, it is necessary to obtain pre-contrast magnetization transfer images in order to accurately identify enhancing lesions on post-contrast magnetization transfer images.⁶¹ Adding two sequences to a conventional protocol requires additional scan time. Similarly, acquiring MTR data to characterize brain pathology necessitates two additional imaging sequences (one with and one without saturation pulses), higher power deposition in tissue, and additional time for data analysis. The authors have used MTR data primarily to measure the severity of white matter lesions in specific genetic diseases and to develop outcome measures in therapy trials. Clinical use of MTR has been very limited.^{59, 60}

Diffusion-weighted imaging (DWI)

DWI detects water molecule random motion, which is affected by various brain lesions, notably infarcts.⁴¹ The apparent diffusion coefficient (ADC) is a quantitative expression of water motion that is calculated from DW images generated using different gradient strengths (b-values). Free water, such as in the cerebral ventricles, has a low signal intensity on DW images and high ADC. In hyperacute infarction, cytotoxic edema appears hyperintense in DW images and hypointense on ADC.^{17, 62, 63} Low ADC correlates with the reduction in the extracellular volume fraction that occurs in infarction as a result of swelling of cells.⁶⁴ As cell swelling resolves and necrosis occurs, ADC increases.^{42, 65} Similar correlations between low ADC and cellular swelling occur following toxin administration,⁶⁶ status epilepticus,⁶⁷ and hypoglycemia.⁶⁸ ADC may also be used to differentiate cytotoxic edema, which decreases extracellular volume, from vasogenic edema, which increases it.⁶⁹ In addition to cell swelling and size of the extracellular space, the mobility of extracellular water appears to be affected by its composition and the presence of cells. ADC has also been used to measure brain maturation;⁴⁹ to differentiate epidermal cysts from arachnoid cysts, gliomas from abscesses, and ependymomas from oligodendrogliomas, and to determine tumor grade and cellularity.⁷⁴

In dogs, DWI and ADC have been used for diagnosis and characterization of acute and chronic ischemic stroke^{17, 18, 70, 71} and as a marker of cytotoxic edema following status epilepticus.⁷² In cats, ADC has been used to detect neuronal swelling associated with experimental reversible ischemia⁴² and naturally-occurring alpha-mannosidosis, a lysosomal storage disease.⁷³ In cats with alpha-mannosidosis, neuronal and glial swelling and astrogliosis occur together and all may contribute to reduced ADC.⁷³

Diffusion tensor imaging (DTI) is a development of DWI that displays directional differences (anisotropy) in water diffusion.^{75, 76} DTI can be used to determine the location and orientation of white matter tracts, a technique known as diffusion tractography. This has proved useful in decreasing post-surgical morbidity,⁷⁴ and characterizing abnormal myelination in multiple sclerosis.⁷⁷ In cats, DTI has been used to study the development of association tracts in the cerebrum,⁷⁸ and to evaluate the development of experimentally-induced vasogenic edema.⁷⁹ Diffusion tractography of the feline cerebrum has also been used to examine the normal post-natal development of the subplate zone and cortical plate into mature cortical pathways.⁸⁰

There are several important practical considerations when using DWI. ^81,82 For example, strong gradients are necessary. Motion of the patient's head with respiration and/or the cardiac cycle may mask water diffusion in the brain and thus require further patient restraint or use of respiratory and/or cardiac gating. Patient temperature affects ADC measurements, hence the patient be maintained at a steady temperature, which for an anaesthetized animal may necessitate blowing warm air through the bore of the magnet. A limitation of diffusion tractography is that the reconstruction algorithms may not accurately display tracts that contain crossing, branching or bending fibers.⁸¹

Magnetic resonance spectroscopy

Magnetic resonance spectroscopy (MRS) may be used to non-invasively evaluate brain biochemistry by quantifying the concentrations of specific metabolites from spectra of metabolite distributions. Proton MRS can be used to detect lactate, creatine (Cr), choline (Cho), phosphocreatine, myoinositol (mI), N-acetylaspartate (NAA), glutamate, and other metabolites⁴⁴ (Figure 6). These metabolites are involved in cellular energy metabolism, cell membrane synthesis, or serve as neuronal markers. For example, NAA is a marker of mature viable neurons (decreases in NAA are indicative of neuronal loss); Cho is a marker of membrane turnover (increases reflect membrane damage affecting myelin or neurons as well as gliosis); mI is a glial cell marker (increases may reflect gliosis); and lactate is a marker of anaerobic metabolism (increases may reflect metabolic abnormalities).⁴⁴

Figure 6.

Proton MRS centered over the centrum semovale (A) of a dog with globoid cell leukodystrophy (GLD). Decreases in NAA (7.4 mM) and increases in choline (4.1 mM) were identified in the affected dog when compared to an unaffected dog (10.6mM and 3.1 mM respectively (B).

Concentrations of metabolites can reveal the biochemistry of specific disease processes including neoplasia,^83-85 cerebral ischemia,⁸⁶ metabolic encephalopathy,⁸⁷ seizures⁸⁸ and neurodegenerative disorders.^{1, 44, 89} In one study of MRS, the pattern of metabolites in a given brain volume was used to correctly classify 104 out of 105 brain tumors of 5 different types.⁸³ MRS has been used to diagnose specific metabolic diseases such as Canavan's disease, which shows elevations in NAA, as well as to monitor response to therapy.^{90, 91} MRS methods have also been developed to detect administered compounds including specific anticancer agents such as fluorouracil, temozolomide and iproplatin^92-94 and to noninvasively monitor the pharmacokinetics of specific anticancer agents.^93-95

In cats with alpha-mannosidosis, in vivo MRS was capable of detecting increased concentrations of the oligosaccharides that are the specific intracellularly-stored substrates associated with this disease.⁹⁶ In FIV-infected cats, in vivo and ex vivo MRS of the brain showed a reduction in both NAA and NAA/Cr ratio, increased trimethylamine/creatine ratio and increased glutamate concentrations in the brain.^97-99 Focal brain ischemia in cats resulted in decreases in both NAA and Cr, and increase in lactate.¹⁰⁰ Cardiac arrest in cats is associated with increased brain lactate concentrations that become elevated within five minutes of arrest and remain high for at least six hours following reperfusion.¹⁰¹

In dogs, MRS has also been used to characterize a number of brain diseases. Increases in glutamate/glutamine and lactate, and decreases in creatine were identified post-ictally.¹⁰² NAA and creatine were indirectly correlated with tumor volume, and lactate was directly correlated with tumor volume in an induced brain tumor study.¹⁰³ Lactate concentrations were associated with intracranial hypertension during fulminant hepatic failure.¹⁰⁴ NAA/Cho ratio was decreased in brain injury due to hypothermic circulatory arrest and the therapeutic efficacy of diazoxide was examined using this measure.¹⁰⁵ Elevations in lactate/creatine and inositol/creatine ratios were used to evaluate cerebral tissue in studying thanatochronology.¹⁰⁶

Practical considerations for performing MRS reflect the balance between maximizing signal-to-noise ratio by increasing voxel size versus achieving spatial accuracy in sampling, which requires a small voxel. Voxel size on 1.5 T scanners may exceed 1 cm³ and inclusion of calvaria within a large voxel cause magnetic field inhomogeneity and can lead to production of uninterpretable spectra. If it is suspected that brain pathology is regional or distributed asymmetrically throughout the brain, chemical shift imaging may be a useful alternative for comparing regional pathology.⁴⁹

Correlating MR findings with clinical signs

Assessment that an MR abnormality is responsible for the clinical signs depends primarily on the degree of correspondence between the site of the lesion and the neuroanatomical localization. Proper neuroanatomical localization of a lesion is based on identification of clinical signs of dysfunction which commonly include 1) cerebral hemisphere/diencephalic dysfunction - altered mental status, circling with preservation of gait, postural reaction deficits, sensory abnormalities, loss of vision, pupillary abnormalities, and seizures, 2) brain stem dysfunction - altered mental status, vestibular and general proprioceptive ataxia, postural reaction deficits, sensory abnormalities, pupillary abnormalities, and cranial nerve dysfunction (excluding the olfactory and optic nerve), and, 3) cerebellar dysfunction – intention tremors, titubation, and cerebellar ataxia. Neurological examination lateralizing signs are also used to predict the lesion side. As a general but imperfect rule, lesions rostral to or including the midbrain are expected to cause contralateral deficits while lesions affecting the pons, medulla oblongata and / or the spinal cord result in ipsilateral deficits.

In many cases it is difficult to know whether an imaging abnormality is responsible for the clinical signs or is an incidental finding. For example, in animals with seizures, but without EEG corroboration of the seizure focus, MR studies at best identify structural or biochemical abnormalities that could result in abnormal electrical activity. Alternatively, some structural abnormalities that may result in neurological dysfunction, are frequently observed in animals without clinical signs. Examples of these include hydrocephalus, caudal occipital malformation syndrome, quadrigeminal cyst, and brain atrophy.

With regard to hydrocephalus, large variations in ventricular volume occur in many dog breeds.¹² Indeed, in both human and veterinary patients hydrocephalus may occur in the absence of readily measureable clinical signs (see “Is your brain really necessary”¹⁰⁷). In this example, which imaging criteria may be used to determine whether hydrocephalus is responsible for the identified clinical signs that may include behavioral abnormalities, altered mental status, ataxia, circling, blindness, or vestibular dysfunction?¹⁰⁸ Although size might be expected to be an important factor, existing measures of ventriculomegaly do not correlate well with the presence of clinical signs.¹⁰⁹ Other patient and physiological variables, including age at which hydrocephalus develops, location of the obstruction (if present), and rate at which the ventricles enlarge, probably reflect better the degree of brain damage and the likelihood of clinical signs.¹¹⁰ One study in children found a correlation between size (area) of the corpus callosum and motor ability and cognitive skills.¹¹¹ Other studies have found some correlations between ventricular size and clinical signs; however, these findings do not enable diagnosis and do not predict outcome following shunt placement.^112,113,114 Diagnosis of clinically significant, normal-pressure hydrocephalus is confirmed only by improvement in clinical signs following shunting of cerebrospinal fluid.¹¹³ A recent study concluded that NAA/Cr and NAA/Cho ratios of the periventricular white matter and phase-contrast MR imaging of cerebrospinal fluid flow at the mesencephalic aqueduct may be useful in predicting which patients could benefit from cerebrospinal fluid shunting.¹¹⁵ Hydrocephalus is one example of many conditions for which there is a need for MR measures that enable more specific diagnosis and prediction of therapeutic outcome.