Enhancing 1p/19q Classification in Brain Gliomas Using Isocitrate Dehydrogenase Status: A Deep Learning Study

Isocitrate dehydrogenase (IDH) mutation and 1p/19q codeletion are critical biomarkers for glioma diagnosis and therapy. 1p/19q codeletion occurs exclusively in IDH-mutated gliomas. In this study, we developed a 2-stage, noninvasive, MRI-based deep learning method that leverages IDH status to enhance 1p/19q predictions.

Multicontrast brain tumor MRI and genomic information were obtained from 5 publicly available (The Cancer Imaging Archive/The Cancer Genome Atlas [TCIA], University of California San Francisco [UCSF], Erasmus Glioma Database [EGD], University of Pennsylvania Health System [UPenn], and Low Grade Glioma [LGG] database) and 3 in-house/collaborator (University of Texas Southwestern Medical Center [UTSW], New York University [NYU], and University of Wisconsin-Madison [UWM]) institutions. Subjects were screened for the availability of IDH and 1p/19q status as well as T1, contrast-enhancing T1, T2, and FLAIR MR images. For training purposes, missing T1 and FLAIR contrasts for the LGG database were generated using an in-house multicontrast simulator. Two separate U-Nets were developed for 1p/19q-classification: a multicontrast network (MC-Net) and a T2-weighted image-only network (T2-Net). A separate U-Net was developed for IDH classification (IDH-Net). A total of 2044 subjects were used in training and testing IDH-Net, and 1426 subjects were used in training and testing the 1p/19q models. The IDH-Net was trained using subjects from TCIA, UTSW, and UPenn. The 1p/19q networks were trained using subjects from TCIA, UTSW, and LGG. The trained networks were tested on true held-out cases from NYU, UWM, EGD, and UCSF. In the 2-stage approach, subjects were initially classified for IDH status using IDH-Net. Predicted IDH wild-type cases default to 1p/19q non-codeleted. Then the IDH-mutated cases were further classified for 1p/19q status using the 1p/19q-networks.

IDH-Net achieved a classification accuracy of 93.7%. 1p/19q MC-Net and T2-Net achieved classification accuracies of 86.5% and 86.0%, respectively. In the 2-stage approach, 1p/19q MC-Net and T2-Net achieved accuracies of 91.5% and 91.2%, respectively, improving the classification accuracy by ∼5%.

This study demonstrates the effectiveness of leveraging IDH status to enhance 1p/19q classification. An ∼5% increase in classification accuracy was achieved when using the 2-stage approach, using IDH-Net to gate 1p/19q predictions. The developed method offers a reliable, noninvasive approach to determine important biomarkers for glioma diagnosis.