Traditional time-frequency analysis is either limited to fixed window widths (short-time Fourier transform) or non-sinusoidal basis functions (wavelets). The recently developed S-transform (ST) overcomes both of these limitations to reveal frequency variation over both space and time. The ST has been used successfully in the medical imaging field for noise removal and texture analysis of brain magnetic resonance (MR) images. We have developed a novel ST-based texture analysis procedure for classifying two genetically distinct forms of a malignant brain tumor known as oligodendroglioma.As only one of these tumor types is sensitive to chemotherapy, correct classification is critical to proper treatment. While traditional genetic analysis requires an invasive biopsy operation, our method is non-invasive, utilizing only information from MR imaging. In this paper we present our non-invasive method for classifying oligodendroglioma tumors using novel STbased texture analysis of MR images as well as a patient study to assess its performance. Our results show a highly significant difference between the two tumor groups on each of three common MR contrasts. Noninvasive classification of tumor genotype could have significant impact in disease management and patient prognosis.