Medulloblastoma is the most common malignant pediatric brain tumor. There have been significant advances in understanding the biology of this tumor, and it has emerged that there are 4 distinct subgroups, each with a different clinical profile and likelihood of survival. The so-called wingless (WNT) group has the best prognosis, and clinical trial designs are now seeking to reduce the burden of therapy for this group; a less aggressive surgical resection and reduction of risk of posterior fossa syndrome or cerebellar mutism may be appropriate. Clearly, this would require presurgical planning based on imaging findings. Advanced MR imaging is increasingly able to characterize these tumors noninvasively through quantitative analysis of images.
Early noninvasive diagnosis and prognosis determination will enhance the planning and refinement of treatment—particularly surgery—as well as inform discussions with family in the crucial early stages of clinical management.
From clinical practice, we know that visual inspection of MRI features allows good classification of posterior fossa tumor type, but not of subtypes. To overcome this limitation, in this work we set out to take morphologic classification to a new level using support vector machine classifiers on quantitative MR imaging features. This paper describes our first attempts to link tumor types and subtypes to features derived from clinical MRI.
The promising results we obtained in this small cohort, with very good discrimination of tumor types and the potential for subtype classification, led us to expand the work to an ongoing, multicenter, multimodal radiogenomics study in medulloblastomas. If confirmed, we will develop a prognostic test for medulloblastoma with the expectation of early clinical benefits that further our understanding of these important tumors.