Object: Data analysis was performed in a multicenter study to evaluate magnetic resonance (MR) imaging for classification of brain tumors, prognosis, and prediction of tumor histological diagnosis.
Methods: The clinical, MR imaging, and histological findings obtained in 142 pediatric cases of brainstem disease were assessed in a multicenter study performed as a blinded review. Clinical data were available in 142 cases, histopathological findings in 126, and MR images in 131. The subgroup of cases involving brainstem gliomas (78 cases) was analyzed separately. Images that met criteria for evaluation were reviewed in random order by three experienced observers who were initially blinded to clinical data as well as histopathological diagnoses. Subsequently, the images were randomized again and provided to the observers for review together with the clinical symptoms of the specific patients. The three observers were able to correctly identify lesions as tumors or nontumorous disease on MR images in 99, 96, and 95% of cases, resulting in an overall sensitivity of 0.94, a specificity of 0.43, a positive predictive value of 0.96, and a negative predictive value of 0.45. Awareness of clinical symptoms did not change the results.
Conclusions: Based on 14 imaging criteria together with the patient's clinical history and symptoms, laboratory data (results of cerebrospinal fluid analysis as well as infectious and immunological parameters), and imaging follow up, a diagnosis of brainstem tumor, as opposed to demyelination, encephalitis, or granuloma, could generally be made. Given these findings, there is only rarely a need for biopsy, and in those patients in whom it is considered, the potential costs and benefits must be carefully assessed on a case-by-case basis.