Dataset Information

MRI-Based Radiomics Combined with Deep Learning for Distinguishing IDH-Mutant WHO Grade 4 Astrocytomas from IDH-Wild-Type Glioblastomas.

ABSTRACT: This study aimed to investigate the potential of quantitative radiomic data extracted from conventional MR images in discriminating IDH-mutant grade 4 astrocytomas from IDH-wild-type glioblastomas (GBMs). A cohort of 57 treatment-naïve patients with IDH-mutant grade 4 astrocytomas (n = 23) and IDH-wild-type GBMs (n = 34) underwent anatomical imaging on a 3T MR system with standard parameters. Post-contrast T1-weighted and T2-FLAIR images were co-registered. A semi-automatic segmentation approach was used to generate regions of interest (ROIs) from different tissue components of neoplasms. A total of 1050 radiomic features were extracted from each image. The data were split randomly into training and testing sets. A deep learning-based data augmentation method (CTGAN) was implemented to synthesize 200 datasets from the training sets. A total of 18 classifiers were used to distinguish two genotypes of grade 4 astrocytomas. From generated data using 80% training set, the best discriminatory power was obtained from core tumor regions overlaid on post-contrast T1 using the K-best feature selection algorithm and a Gaussian naïve Bayes classifier (AUC = 0.93, accuracy = 0.92, sensitivity = 1, specificity = 0.86, PR_AUC = 0.92). Similarly, high diagnostic performances were obtained from original and generated data using 50% and 30% training sets. Our findings suggest that conventional MR imaging-based radiomic features combined with machine/deep learning methods may be valuable in discriminating IDH-mutant grade 4 astrocytomas from IDH-wild-type GBMs.

SUBMITTER: Hosseini SA

PROVIDER: S-EPMC9913426 | biostudies-literature | 2023 Feb

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

MRI-Based Radiomics Combined with Deep Learning for Distinguishing IDH-Mutant WHO Grade 4 Astrocytomas from IDH-Wild-Type Glioblastomas.

Hosseini Seyyed Ali SA Hosseini Elahe E Hajianfar Ghasem G Shiri Isaac I Servaes Stijn S Rosa-Neto Pedro P Godoy Laiz L Nasrallah MacLean P MP O'Rourke Donald M DM Mohan Suyash S Chawla Sanjeev S

Cancers 20230202 3

This study aimed to investigate the potential of quantitative radiomic data extracted from conventional MR images in discriminating IDH-mutant grade 4 astrocytomas from IDH-wild-type glioblastomas (GBMs). A cohort of 57 treatment-naïve patients with IDH-mutant grade 4 astrocytomas (<i>n</i> = 23) and IDH-wild-type GBMs (<i>n</i> = 34) underwent anatomical imaging on a 3T MR system with standard parameters. Post-contrast T1-weighted and T2-FLAIR images were co-registered. A semi-automatic segment ...[more]

PMID: 36765908

Similar Datasets

Project description:Telomerase reverse transcriptase (TERT) promoter (pTERT) mutation has often been described as a late event in gliomagenesis and it has been suggested as a prognostic biomarker in gliomas other than 1p19q codeleted tumors. However, the characteristics of isocitrate dehydrogenase (IDH) wild type (wt) (IDHwt), pTERTwt glioblastomas are not well known. We recruited 72 adult IDHwt, pTERTwt glioblastomas and performed methylation profiling, targeted sequencing, and fluorescence in situ hybridization (FISH) for TERT structural rearrangement and ALT (alternative lengthening of telomeres). There was no significant difference in overall survival (OS) between our cohort and a the Cancer Genome Atlas (TCGA) cohort of IDHwt, pTERT mutant (mut) glioblastomas, suggesting that pTERT mutation on its own is not a prognostic factor among IDHwt glioblastomas. Epigenetically, the tumors clustered into classic-like (11%), mesenchymal-like (32%), and LGm6-glioblastoma (GBM) (57%), the latter far exceeding the corresponding proportion seen in the TCGA cohort of IDHwt, pTERTmut glioblastomas. LGm6-GBM-clustered tumors were enriched for platelet derived growth factor receptor alpha (PDGFRA) amplification or mutation (p = 0.008), and contained far fewer epidermal growth factor receptor (EGFR) amplification (p < 0.01), 10p loss (p = 0.001) and 10q loss (p < 0.001) compared with cases not clustered to this group. LGm6-GBM cases predominantly showed ALT (p = 0.038). In the whole cohort, only 35% cases showed EGFR amplification and no case showed combined chromosome +7/-10. Since the cases were already pTERTwt, so the three molecular properties of EGFR amplification, +7/-10, and pTERT mutation may not cover all IDHwt glioblastomas. Instead, EGFR and PDGFRA amplifications covered 67% and together with their mutations covered 71% of cases of this cohort. Homozygous deletion of cyclin dependent kinase inhibitor 2A (CDKN2A)/B was associated with a worse OS (p = 0.031) and was an independent prognosticator in multivariate analysis (p = 0.032). In conclusion, adult IDHwt, pTERTwt glioblastomas show epigenetic clustering different from IDHwt, pTERTmut glioblastomas, and IDHwt glioblastomas which are pTERTwt may however not show EGFR amplification or +7/-10 in a significant proportion of cases. CDKN2A/B deletion is a poor prognostic biomarker in this group.

Project description:IDH-mutant astrocytomas have a more indolent natural history and better prognosis than their IDH-wild type counterparts, but are still graded according to schemes developed prior to the recognition of this type of neoplasm as a distinct entity. Homozygous deletion of CDKN2A has been proposed as a molecular correlate of aggressive behavior in these tumors, and may be incorporated into future grading systems in an effort to improve prognostic stratification. Fluorescence in situ hybridization (FISH) is a common ancillary testing modality used to assess CDKN2A status, but the specifics of how to best interpret FISH results for prognostication of gliomas have not been clearly defined in the literature. To address this issue, we performed a retrospective analysis of prospectively collected CDKN2A FISH data from 108 primary and 43 recurrent IDH-mutant astrocytomas diagnosed between 2007-2020 at the University of Pittsburgh Medical Center. High level CDKN2A homozygous deletion was rare in primary tumors and was identified more frequently in recurrent tumors. Multivariate Cox Proportional-Hazards analysis demonstrated that histologic grade and CDKN2A status are independent predictors of survival, and the prognostic value of CDKN2A is maximized by applying a threshold of ≥ 30% of tumor cells with homozygous deletion by FISH to define a positive result. At this threshold, CDKN2A deletion significantly stratified survival of histologic grade 4 tumors, but grade 2 and 3 tumors rarely exceeded this cutoff value and did not show worse survival. Lower thresholds identified additional lower grade tumors, but were not prognostically useful. Compared to prior studies, the lack of prognostic significance of CDKN2A homozygous deletion by FISH in grade 2-3 IDH-mutant astrocytomas may reflect differences in cohort populations or technical differences between testing modalities. Definitive criteria for determining CDKN2A homozygous deletion by various methodologies will be critical if this is to be included in future grading schemes.

Dataset Information

MRI-Based Radiomics Combined with Deep Learning for Distinguishing IDH-Mutant WHO Grade 4 Astrocytomas from IDH-Wild-Type Glioblastomas.

Publications

MRI-Based Radiomics Combined with Deep Learning for Distinguishing IDH-Mutant WHO Grade 4 Astrocytomas from IDH-Wild-Type Glioblastomas.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets