AwardsStudent Travel Stipend Award
ParticipantsSevcan Turk, MD, izmir, Turkey (Presenter) Nothing to Disclose
Kaya Oguz, Izm?r, Turkey (Abstract Co-Author) Nothing to Disclose
Omer Kitis, Izmir, Turkey (Abstract Co-Author) Nothing to Disclose
Cem Calli, Izmir, Turkey (Abstract Co-Author) Nothing to Disclose
Cenk Eraslan, MD, Izmir, Turkey (Abstract Co-Author) Nothing to Disclose
Mehmet N. Orman, Izmir, Turkey (Abstract Co-Author) Nothing to Disclose
Taner Akalin, Izmir, Turkey (Abstract Co-Author) Nothing to Disclose
Taskin Yurtseven, Izm?r, Turkey (Abstract Co-Author) Nothing to Disclose
sevcanturk.ege@hotmail.com
PURPOSEThe purpose of this study is to classify glial tumors into grade II, III and IV categories noninvasively by application of machine learning to multi-modal MRI features.
METHOD AND MATERIALSWe retrospectively studied 57 glioma patients with pre-postcontrast T1, T2, FLAIR, ADC maps which were acquired 3T MRI. Two radiologists segmented the tumors into enhancing and nonenhancing part, tumor necrosis, cyst and edema using semi-automated segmentation tools. We measured total tumor volume, enhancing and nonenhancing tumor volume, edema volume, tumor necrosis volume and the ratios to the total volume for each image. Training of a support vector machine (SVM) classifier was performed with labeled data designed to answer the question of interest. Specificity, sensitivity, and AUC of the predictions were computed by means of ROC analysis. Differences in continuous measures between groups were assessed by using Kruskall Wallis, with post hoc Dunn correction for multiple comparisons.
RESULTSWhen we compared the volume ratios between groups there was statistically significant difference between grade IV and grade II-III glial tumors. Edema and tumor necrosis volume ratios for grade IV glial tumors were higher than that of grade II and III. Volumetric ratio analysis could not distinguish grade II and III tumors. However, SVM correctly classified each group with accuracies up to 93%.
CONCLUSIONRadiomics and machine learning are emerging techniques that extract unrevealed information from medical images. Application of machine learning methods to MRI features can be used to classify tumors noninvasively and more readily in clinical settings.
CLINICAL RELEVANCE/APPLICATIONThe large amount of data produced by MRI limits the use of precise quantitative measurements in the clinical practice. Therefore, automated and reliable machine learning methods are required. Application of machine learning methods to MRI features can be used to classify tumors noninvasively and to predict prognostic information.Our study will be helpful to extract radiomic information from MRI features that are useful for making predictions noninvasively and for making classifications more readily in clinical settings.