Radiomics-Based Machine Learning in Differentiation Between Glioblastoma and Metastatic Brain Tumors

The objective of this study is to investigate the use of pattern classification methods for distinguishing different types of brain tumors, such as primary gliomas from metastases, and also for grading of gliomas. The availability of an automated computer analysis tool that is more objective than human readers can potentially lead to more reliable and reproducible brain tumor diagnostic procedures. A computer-assisted classification method combining conventional MRI and perfusion MRI is developed and used for differential diagnosis. The proposed scheme consists of several steps including region-of-interest definition, feature extraction, feature selection, and classification. The extracted features include tumor shape and intensity characteristics, as well as rotation invariant texture features. Feature subset selection is performed using support vector machines with recursive feature elimination. The method was applied on a population of 102 brain tumors histologically diagnosed as metastasis (24), meningiomas (4), gliomas World Health Organization grade II (22), gliomas World Health Organization grade III (18), and glioblastomas (34). The binary support vector machine classification accuracy, sensitivity, and specificity, assessed by leave-one-out cross-validation, were, respectively, 85%, 87%, and 79% for discrimination of metastases from gliomas and 88%, 85%, and 96% for discrimination of high-grade (grades III and IV) from low-grade (grade II) neoplasms. Multiclass classification was also performed via a one-vs-all voting scheme. (c) 2009 Wiley-Liss, Inc.

To determine whether perfusion-weighted and proton spectroscopic MR imaging can be used to differentiate high-grade primary gliomas and solitary metastases on the basis of differences in vascularity and metabolite levels in the peritumoral region. Fifty-one patients with a solitary brain tumor (33 gliomas, 18 metastases) underwent conventional, contrast material--enhanced perfusion-weighted, and proton spectroscopic MR imaging before surgical resection or stereotactic biopsy. Of the 33 patients with gliomas, 22 underwent perfusion-weighted MR imaging; nine, spectroscopic MR imaging; and two underwent both. Of the 18 patients with metastases, 12 underwent perfusion-weighted MR imaging, and six, spectroscopic MR imaging. The peritumoral region was defined as the area in the white matter immediately adjacent to the enhancing (hyperintense on T2-weighted images, but not enhancing on postcontrast T1-weighted images) portion of the tumor. Relative cerebral blood volumes in these regions were calculated from perfusion-weighted MR data. Spectra from the enhancing tumor, the peritumoral region, and normal brain were obtained from the two-dimensional spectroscopic MR acquisition. The Student t test was used to determine if there was a statistically significant difference in relative cerebral blood volume and metabolic ratios between high-grade gliomas and metastases. The measured relative cerebral blood volumes in the peritumoral region in high-grade gliomas and metastases were 1.31 +/- 0.97 (mean +/- SD) and 0.39 +/- 0.19, respectively. The difference was statistically significant (P <.001). Spectroscopic imaging demonstrated elevated choline levels (choline-to-creatine ratio was 2.28 +/- 1.24) in the peritumoral region of gliomas but not in metastases (choline-to-creatine ratio was 0.76 +/- 0.23). The difference was statistically significant (P =.001). Although conventional MR imaging characteristics of solitary metastases and primary high-grade gliomas may sometimes be similar, perfusion-weighted and spectroscopic MR imaging enable distinction between the two.

Diffusion-weighted magnetic resonance (MR) imaging and perfusion MR imaging are advanced techniques that provide information not available from conventional MR imaging. In particular, these techniques have a number of applications with regard to characterization of tumors and assessment of tumor response to therapy. In this review, the authors describe the fundamental principles of diffusion-weighted and perfusion MR imaging and provide an overview of the ways in which these techniques are being used to characterize tumors by helping distinguish tumor types, assess tumor grade, and attempt to determine tumor margins. In addition, the role of these techniques for evaluating response to tumor therapy is outlined.