Prediction of survival in patients affected by glioblastoma: histogram analysis of perfusion MRI

Romano, Andrea; Pasquini, Luca; Napoli, A. Di; Tavanti, Francesca; Boellis, Alessandro; Espagnet, Maria Camilla Rossi; Minniti, Giuseppe; Bozzao, Alessandro

doi:10.1007/s11060-018-2887-4

Cited by 20 publications

(13 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The high IDH prediction accuracy obtained with perfusion images is consistent with previous studies [ 41 , 42 ]. Besides reflecting tumor grade [ 43 ], perfusion parameters showed a promising correlation with patient survival, particularly rCBV [ 44 ]. In our experience, we achieved high predictivity of GBM IDH status with machine ML on rCBV data [ 28 ].…”

Section: Discussionmentioning

confidence: 99%

Deep Learning Can Differentiate IDH-Mutant from IDH-Wild GBM

Pasquini

Napolitano

Tagliente

et al. 2021

JPM

Self Cite

View full text Add to dashboard Cite

Isocitrate dehydrogenase (IDH) mutant and wildtype glioblastoma multiforme (GBM) often show overlapping features on magnetic resonance imaging (MRI), representing a diagnostic challenge. Deep learning showed promising results for IDH identification in mixed low/high grade glioma populations; however, a GBM-specific model is still lacking in the literature. Our aim was to develop a GBM-tailored deep-learning model for IDH prediction by applying convoluted neural networks (CNN) on multiparametric MRI. We selected 100 adult patients with pathologically demonstrated WHO grade IV gliomas and IDH testing. MRI sequences included: MPRAGE, T1, T2, FLAIR, rCBV and ADC. The model consisted of a 4-block 2D CNN, applied to each MRI sequence. Probability of IDH mutation was obtained from the last dense layer of a softmax activation function. Model performance was evaluated in the test cohort considering categorical cross-entropy loss (CCEL) and accuracy. Calculated performance was: rCBV (accuracy 83%, CCEL 0.64), T1 (accuracy 77%, CCEL 1.4), FLAIR (accuracy 77%, CCEL 1.98), T2 (accuracy 67%, CCEL 2.41), MPRAGE (accuracy 66%, CCEL 2.55). Lower performance was achieved on ADC maps. We present a GBM-specific deep-learning model for IDH mutation prediction, with a maximal accuracy of 83% on rCBV maps. Highest predictivity achieved on perfusion images possibly reflects the known link between IDH and neoangiogenesis through the hypoxia inducible factor.

show abstract

Section: Discussionmentioning

confidence: 99%

Deep Learning Can Differentiate IDH-Mutant from IDH-Wild GBM

Pasquini

Napolitano

Tagliente

et al. 2021

JPM

Self Cite

View full text Add to dashboard Cite

show abstract

“…The prognosis potential of the rCBV in glioblastoma at the ET ROI has been extensively studied 14,34,35 . The prognosis potential of the rCBV in glioblastoma in the non-enhancing region was also suggested in some studies 15,23,36 .…”

Section: Discussionmentioning

confidence: 99%

Higher vascularity at infiltrated peripheral edema differentiates proneural glioblastoma subtype

Chelebian

Fuster–Garcia

Álvarez‐Torres

et al. 2020

Preprint

View full text Add to dashboard Cite

2 14 ABSTRACT 15 BACKGROUND AND PURPOSE: Genetic classifications are crucial for 16 understanding the heterogeneity of glioblastoma. Recently, MR perfusion imaging 17 techniques have demonstrated their ability to determine molecular alterations. In this 18 work, we investigated whether perfusion markers within infiltrated peripheral edema 19 were associated with proneural, mesenchymal, classical and neural subtypes.20 MATERIALS AND METHODS: ONCOhabitats open web service was used to obtain 21 the cerebral blood volume at the infiltrated peripheral edema for MRI studies of 50 22 glioblastoma patients from The Cancer Imaging Archive: TCGA-GBM. ANOVA and 23 Kruskal-Wallis tests were carried out in order to assess the association between vascular 24 features and the subtypes. For assessing specific differences, Mann-Whitney U-test was 25 conducted. Finally, the association of overall survival with molecular and vascular 26 features was assessed using univariate and multivariate Cox models. 27 RESULTS: ANOVA and Kruskal-Wallis tests for the maximum cerebral blood volume 28 at the infiltrated peripheral edema between the four subclasses yielded false discovery 29 rate corrected p-values of <0.001 and 0.02, respectively. This vascular feature was 30 significantly higher (p=0.0043) in proneural patients compared to the rest of the subtypes 31 while conducting Mann-Whitney U-test. The multivariate Cox model pointed to 32 redundant information provided by vascular features at the peripheral edema and 33 proneural subtype when analyzing overall survival.34 CONCLUSIONS: Higher relative cerebral blood volume at infiltrated peripheral edema 35 is associated with proneural glioblastoma subtype suggesting underlying vascular 36 behavior related to molecular composition in that area. 3 37 1. INTRODUCTION 38 In the late years, Central Nervous System tumor classification has shifted from being 39 based on microscopic similarities between cells and their levels of differentiation 1 to 40 additionally include genetic-based features 2 . This is particularly the case for glioblastoma, 41 where several classifications have been defined: on the one hand, the World Health 42 Organization (WHO) classification which distinguishes between IDH-wildtype and IDH-43 mutant glioblastomas 2-4 and, on the other, the Verhaak classification 5 , consisting of 4 44 subtypes depending on mutations and molecular profile of various cancer-related genes. 45 These subtypes are the mesenchymal, classical, neural and proneural, the latter being 46 related to IDH mutations 5,6 . These new classification paradigms have improved the 47 estimation of prognosis 7,8 and proposed specific therapeutic targets 9-12 , especially for 48 patients with proneural and mesenchymal type glioblastoma. 49 Considering that Magnetic Resonance Imaging (MRI) perfusion biomarkers have been 50 associated with patients' overall survival 13-15 and cellular features 16,17 , several studies 51 were performed to analyze if there was a relationship between vascular biomarkers and 52 the gen...

show abstract

“…Histogram parameters can reflect the microstructure of tissues by issuing every voxel of the MRI into a histogram, and thus, statistical information of the images can be obtained 6‐12 . Thus, in oncologic imaging, it was extensively researched that histogram parameters can reflect microstructure in tumors, which was shown for several tumor entities 6‐12 . There are several different parameters, which can be calculated.…”

Section: Discussionmentioning

confidence: 99%

“…Previously, some preliminary studies, predominantly in the field of oncologic imaging, showed that histogram parameters derived from MRI are capable to reflect different microstructure characteristics of tumors, such as cellularity and proliferation potential 7‐10 . Furthermore, histogram analysis parameters were able to predict treatment response in longitudinal analyses in several malignancies 6,11,12 . So, it was discussed that even morphological sequences can reflect tissue features, comprising proliferation potential, cellularity, and nucleic sizes in several malignant tumors 7‐10 …”

Section: Introductionmentioning

confidence: 99%