HGG-35. COMBINATIONS OF QUANTITATIVE AND QUALITATIVE MRI FEATURES IDENTIFY PROGNOSTIC AND MOLECULAR SUBGROUPS OF SUPRATENTORIAL PEDIATRIC HIGH-GRADE GLIOMA. Issue 2 (22nd June 2018)

• Record Type: Journal Article
• Authors: Lucas, John; Hsu, Chih-Yang; Lu, Zhaohua; Becksfort, Jared; Tinkle, Christopher; Broniscer, Alberto; Merchant, Thomas; Orr, Brent; Baker, Suzanne; Patay, Zoltan; Hwang, Scott
• Is Part Of: Neuro-oncology. Volume 20:Issue 2(2018)supplement 2
• Page Start: i96
• Page End: i96
• Publication Date: 2018-06-22
• Subjects: Brain Neoplasms -- Periodicals; Brain -- Tumors -- Periodicals; Brain -- Cancer -- Periodicals; Nervous system -- Cancer -- Periodicals; 616.99481
• DOI: 10.1093/neuonc/noy059.307
• Languages: English
• ISSNs: 1522-8517

Abstract

Abstract: BACKGROUND: To evaluate the potential of quantitative (radiomic) and qualitative (Visually Accessible Rembrandt Imaging, VASARI) imaging features to distinguish between molecularly defined subgroups in supratentorial pediatric high grade glioma (pHGG) and improve prognostication. METHOD AND MATERIALS: Eighty-eight consecutive cases of newly diagnosed supratentorial pHGG with complete pre-operative imaging were systematically reviewed and scored for VASARI and radiomic features. Agreement between two expert reviewers was scored for VASARI features. Calculated radiomic features describing intensity, shape, and texture were extracted from T1, T1+Gd, T2, and FLAIR MRI sequences. Hierarchical clustering is used with the distance based on Spearman correlation and complete-linkage. Bootstrapped patient clusters were verified by reviewing the consensus matrix from bootstrapped samples. Subgroup specific survival was related to pHGG defining mutations, fusions, and amplifications. All analyses were completed in either SAS v9.3 or R 3.3.3. RESULTS: VASARI features were variably concordant across reviewers (median 60% (range, 29-79%)). Univariate cox proportional hazards analysis of VASARI features identified deep white matter invasion (HR 4.1 95% CI 1.9-8.9, p<0.001), multi-centric disease (HR 2.5 95% CI 1.2-5.3, p=0.02), lack of calvarial remodelling (HR 4.5 95% CI 1.6-12.5, p=0.004), and cerebellar invasion (HR 4.5 95% CI 1.7-12.1, p=0.003) as features which increased the hazard for death. Hierarchical clustering of radiomic features identified 2 dominant and 1 lesser patient subgroup with distinctive imaging features, histologic identity and differential survival. Evaluation of the relationship between imaging defined patient subgroups and molecular features is ongoing. CONCLUSION: Quantitative and qualitative imaging features refine patient subgrouping and improve prognostication.