Predicting the molecular subtypes in gliomas using T2-weighted mri

Abstract

Classifying gliomas noninvasively into their molecular subsets is a crucial neuro-scientific problem. Prognosis of the isocitrate dehydrogenase (IDH) mutation in gliomas is important for planning targeted therauptic intervention and tailored treatment for individual patients. This work proposes a novel technique based on texture analysis of T2-weighted magnetic resonance imaging (MRI) scans of grade 2 and grade 3 gliomas to differentiate between IDH1 mutant 1p/19q positive and IDH1 mutant 1p/19q negative categories. The textural features used in the proposed method are local binary patterns histogram (LBPH), Shannon entropy, histogram, skewness, kurtosis, and intensity grading. We discriminate the tumors into their molecular subtypes using standard artificial neural networks (ANNs). LBPH attributes demonstrated maximum discrimination between the two groups followed by Shannon entropy. In summary, the technique proposed facilitates an early biomarker to detect the IDH subtype noninvasively and can be employed as an automated tool in clinics to aid diagnosis.