Problem

A 12-year-old spayed female mixed-breed dog was brought to the University of Georgia Veterinary Teaching Hospital for evaluation because of 2 clusters of generalized seizures. Four months previously, the dog had had 5 generalized seizures in a single day. One week before the evaluation, the dog had had a second episode of cluster seizures (6 generalized seizures in a 24 hour-period).

At the time of hospital admission, a physical examination and neurologic examination were performed, and results were unremarkable. The neurologic disturbance was localized to the prosencephalic region on the basis of the dog's history of seizure activity. A CBC, serum biochemical analyses, and thoracic radiography failed to reveal any clinically important abnormalities. Therefore, the differential diagnoses considered were reactive seizures (extracranial disorders), so-called symptomatic epilepsy (intracranial disorders), idiopathic (primary) epilepsy, and so-called probable symptomatic (cryptogenic) epilepsy. Extracranial causes such as metabolic or toxic disorders were considered but did not seem likely given the unremarkable clinicopathologic data and findings during interictal examinations. Idiopathic epilepsy appeared unlikely because the dog was 12 years of age, and this disease is typically first identified when dogs are between 6 months and 5 years of age. Intracranial diseases, such as degenerative conditions, anomalies, neoplasia, inflammatory conditions, trauma, or vascular injury, were considered possible. Neoplasia was considered most likely because seizures are the most common clinical sign in dogs > 5 years of age with brain tumors.¹ Cryptogenic epilepsy was also considered as a possible cause. Consequently, magnetic resonance imaging (MRI) of the brain was performed with a 3.0-Tesla unit.^a

Imaging revealed a large, well-defined, extra-axial ovoid cystic or fluid-filled lesion within the right frontal and olfactory lobes (Figure 1). The lesion had the following MRI characteristics: isointense relative to the lateral ventricles on T2-weighted (T2W) images, hypointense on T1-weighted (T1W) fluid-attenuated inversion recovery (FLAIR) images, and partially suppressed on T2W FLAIR images relative to the surrounding brain tissue. Within the rostroventral aspect of the lesion, there was a well-circumscribed mass that was iso- to hyperintense relative to the surrounding brain tissue on T2W images and isointense on T1W FLAIR images. The mass had mild, heterogeneous enhancement on T1W FLAIR images obtained after IV administration of contrast medium.^b It was associated with a deviation of the falx cerebri to the left side. The internal capsule of the right cerebrum was hyperintense relative to the surrounding white matter on T2W and T2W FLAIR images, consistent with perilesional edema.

Sagittal T-2 weighted magnetic resonance image of the brain of a dog examined because of cluster seizures. Notice the hyperintense mass in the olfactory and frontal lobes with a large cystic or fluid-filled caudal extension (asterisk).

Citation: Journal of the American Veterinary Medical Association 239, 1; 10.2460/javma.239.1.60

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Sagittal T-2 weighted magnetic resonance image of the brain of a dog examined because of cluster seizures. Notice the hyperintense mass in the olfactory and frontal lobes with a large cystic or fluid-filled caudal extension (asterisk).

Citation: Journal of the American Veterinary Medical Association 239, 1; 10.2460/javma.239.1.60

• Download Figure
• Download figure as PowerPoint slide

Sagittal T-2 weighted magnetic resonance image of the brain of a dog examined because of cluster seizures. Notice the hyperintense mass in the olfactory and frontal lobes with a large cystic or fluid-filled caudal extension (asterisk).

Citation: Journal of the American Veterinary Medical Association 239, 1; 10.2460/javma.239.1.60

• Download Figure
• Download figure as PowerPoint slide

Formulation of the Clinical Question

The clinical question was formulated on the basis of an initial assumption that the lesion observed during MRI represented an intracranial neoplasm. This assumption was the product of clinical experience of the authors, the MRI characteristics of the lesion, and the age and clinical signs of the dog.

In general, treatment for intracranial neoplasia in dogs involves palliative or definitive treatments such as surgery, radiation therapy, and chemotherapy alone or in various combinations. Histologic tumor type dictates treatment and prognosis; however, determining tumor type requires histologic examination of a biopsy sample. Brain biopsy is an invasive and expensive procedure unless performed at the time of tumor resection. Additionally, image-guided biopsy of brain lesions is not widely available in veterinary medicine and so a less invasive, more widely available diagnostic tool such as MRI would be a more attractive option if it could reliably yield accurate information on histologic type.

Clinical Question

In dogs with intracranial tumors, can histologic tumor type be accurately identified through MRI alone?

Evidentiary Search Strategy

An evidence-based approach to the clinical question was desired. A targeted literature search was performed through the PubMed and Web of Science databases with the following keywords: magnetic resonance imaging, brain tumors, diagnosis, and dogs. This search identified 130 publications that were evaluated for their relevancy to the clinical question. To be considered further, studies were required to be original research focusing on imaging characteristics of brain tumors, include histologic confirmation of tumor type, and include characteristics distinguishing tumors from other types of lesions and comparisons of characteristics among tumor types. The associated reports were also required to be written in English. Use of these criteria resulted in identification of 8 scientific reports^2–9 as relevant.

Review of Evidence

The evidentiary value of the articles was evaluated on the basis of study design when possible. Given the literature at hand on diagnostic test evaluation, prospective blinded studies involving pathological evaluations of multiple tumor types were deemed to have the highest evidentiary value; retrospective studies involving multiple tumor types and their imaging characteristics were considered the next most valuable, with slightly more value than retrospective studies involving only 1 tumor type; and case reports were deemed to have the least value. Most reports concerned retrospective studies^3–9 in which historical pathologist reports of examinations of necropsy or surgical biopsy specimens were reviewed. One case report,² in which 2 dogs were evaluated, was also included.

Three of the reports^5,7,9 involving retrospective evaluation studies and 1 case report² described MRI findings and histopathologic characteristics of the following specific tumor types: meningioma, glioblastoma multiforme, choroid plexus, and cystic menigiomas. The remaining reports^3,4,6,8 included discussion of the histopathologic characteristics and comparative MRI findings for multiple tumor types. A summary of the MRI characteristics of these tumors is listed (Table 1).

Through multivariate logistic regression, investigators in 1 retrospective evaluation study³ found 7 MRI characteristics to be significantly associated with neoplasia versus lesions of nonneoplastic etiologies: a single lesion, a regular shape, mass effect, dural contact, dural tail, adjacent bone affected, and contrast enhancement. The MRI of the dog in the present case had 5 of the 7 signs associated with neoplasia, including a single lesion, a regular shape, mass effect, dural contact, and contrast enhancement. The presence of a cystic lesion, which was also present in the MRI of the dog in the present case, had a positive predictive value of 91% for intracranial tumors in that study.

The strength of association between MRI findings and tumor type was poorly reported in the literature reviewed. Associations have been drawn by calculation of the percentage of tumors correctly predicted on the basis of pathological characteristics. A different retrospective study⁶ involving evaluation of 24 canine MRI studies found radiologists were able to accurately predict neoplasia versus other pathological processes in 100% of cases. However, the specific histologic tumor type was accurately predicted only 71% of the time. The data in that report were retrieved from medical records rather than through reevaluation of archived MRIs by observers blinded to dog identity. The study also lacked specific criteria for tumor identification and determination of specific tumor types. An earlier retrospective evaluation study⁸ found that gliomas (n = 11), choroid plexus papillomas (5), and pituitary adenomas (3) were diagnosed correctly via MRI. In that study, interpretation of MRI findings was performed by 2 observers who were only aware that they were evaluating histologically confirmed brain tumors and who used objective descriptive criteria for tumor type determination. Meningiomas were diagnosed correctly in 7 of 8 dogs; 1 tumor was incorrectly diagnosed as a glioma because of its cystic appearance and intra-axial location.

A larger retrospective evaluation study⁷ involved 112 dogs with histologic confirmation of intracranial meningioma. In that study, MRIs were classified on the basis of a consensus opinion of 2 neurologists and 1 radiologist, who used objective criteria (including signal intensity and tumor border definition). The following data were obtained: 89% of meningiomas had heterogenous T1W signal intensity with 69% being isointense, 69% had homogeneous T2W signal intensity, 66% had homogeneous contrast enhancement, and 87% had sharply defined borders. Peritumoral edema was associated with 43% (48/112) of meningiomas, and cysts were detected in 26% (29/112). That study found no significant associations between tumor subtypes or grade and any of the MRI features evaluated.

The MRI characteristics of glioblastoma multi-forme, categorized by the World Health Organization as grade IV astrocytoma, were evaluated in a pathological retrospective evaluation study⁵ involving 5 dogs. The MRI information was obtained from patient records. Peritumoral edema and a mass effect were observed in all 5 dogs, well-defined borders were observed in 4 dogs, and ring (peripheral) enhancement on T1W images following contrast administration was observed in 3 dogs. Heterogeneous signal intensity on T2W images (5/5) and isointense-to-hypointense signals on T1W images (5/5) were noted. Two of the glioblastomas in that study had an associated cystic structure.

The MRI characteristics of histologically confirmed choroid plexus tumors have also been described.^4,6,8,9 The most robust study,⁹ a retrospective case series, involved investigation of associations between tumor MRI characteristics and confirmed pathological findings in 56 dogs. Blinded review of the MRIs was carried out by a neurologist and a radiologist, and final categorization of objective characteristics (including signal intensity and uniformity) was achieved through a consensus of opinion. Sixty-three percent of the choroid plexus tumors had peritumoral edema, 75% were associated with ventriculomegaly, 82% were iso- to hyperintense on T1W images, 97% were enhanced with contrast medium administration, and 87% were hyperintense on T2W images.

Given the aforementioned evidence, what decision would you make?

Clinical Decision and Outcome

On the basis of the MRI characteristics of the intracranial lesion identified in the dog reported here combined with the authors' experience and the dog's signalment and history of seizures, a presumptive diagnosis of a menigioma was made. This presumptive diagnosis was discussed with the owners in addition to the treatment strategies and prognosis associated with this tumor type in dogs.

The owners opted for surgical resection of the mass, which was subsequently submitted for histologic assessment. A histopathologic diagnosis of menigioma was made. The owners were prepared for this diagnosis and had already considered additional treatment options; however, none were pursued. The dog continued to receive prednisone and anticonvulsants. Three months following surgery, MRI was repeated, revealing signal characteristics within the white matter of the olfactory and frontal lobes of the cerebrum compatible with edema. Tumor recurrence was not evident. The dog was lost to follow-up 10 months after the surgery.

Typically, when designing a study to evaluate the usefulness of diagnostic tests, there are 2 options to consider: use of historical data, even though results may not be available for reinterpretation and checking for accuracy, and use of prospectively obtained data, which allows a more standardized approach to data collection and interpretation. Thus, the range of study designs available for consideration of quality of evidence is limited. The studies used as evidence in the present clinical scenario were limited by the low number of subjects, differences in disease duration and treatment prior to examination, and absence of MRI characterization of some intracranial tumor types. In addition, studies that are retrospective in nature, such as those evaluated here, are typically viewed as providing lower quality evidence than prospectively conducted studies. To the authors' knowledge, no prospective blinded studies have been performed to evaluate the association between MRI diagnosis and histologic diagnosis in veterinary medicine. This does not render the information available of little use. Rather, in circumstances such as ours, the information available can be helpful in decision making, particularly when considered as a whole.

It is important for practitioners and specialists in neurology, oncology, radiation therapy, and radiology to be aware of the aforementioned limitations when assessing the veterinary literature for answers to clinical questions. However, descriptions based on retrospective information can and did assist in prioritization of differential diagnoses.