The effect of glioblastoma heterogeneity on survival stratification: a multimodal MR imaging texture analysis

Abstract

Background Quantitative evaluation of the effect of glioblastoma (GBM) heterogeneity on survival stratification would be critical for the diagnosis, treatment decision, and follow-up management. Purpose To evaluate the effect of GBM heterogeneity on survival stratification, using texture analysis on multimodal magnetic resonance (MR) imaging. Material and Methods A total of 119 GBM patients (65 in long-term and 54 in short-term survival group, separated by overall survival of 12 months) were selected from the Cancer Genome Atlas, who underwent the T1-weighted (T1W) contrast-enhanced (CE), T1W, T2-weighted (T2W), and FLAIR sequences. For each sequence, the co-occurrence matrix, run-length matrix, and histogram features were extracted to reflect GBM heterogeneity on different scale. The recursive feature elimination based support vector machine was adopted to find an optimal subset. Then the stratification performance of four MR sequences was evaluated, both alone and in combination. Results When each sequence used alone, the T1W-CE sequence performed best, with an area under the receiver operating characteristic curve, accuracy, sensitivity, and specificity of 0.7915, 80.67%, 78.45%, and 83.33%, respectively. When the four sequences combined, the stratification performance was basically equal to that of T1W-CE sequence. In the optimal subset of features extracted from multimodality, those from the T2W sequence weighted the most. Conclusion All the four sequences could reflect heterogeneous distribution of GBM and thereby affect the survival stratification, especially T1W-CE and T2W sequences. However, the stratification performance using only the T1W-CE sequence can be preserved with omission of other three sequences, when investigating the effect of GBM heterogeneity on survival stratification.

A Multi-parametric MRI-Based Radiomics Signature and a Practical ML Model for Stratifying Glioblastoma Patients Based on Survival Toward Precision Oncology

“…The results comparison of the proposed method (signature model and the practical ML model) with most relevant published studies are presented in Table 5. While the proposed method's results, the signature model and the ML model, was not impressive compared to most recently reported works (Macyszyn et al, 2016; Liu et al, 2018b; Sanghani et al, 2018; Chen et al, 2019), it was comparable or even better with respect to others studies for example that reported by Yang et al (2015) (AUC = 0.67 for 12 months survival prediction) and Chaddad et al (2019b) (AUC = 0.78 for short- vs. long-term OS prediction). Also, this study results are relatively comparable with that obtained by Zong et al (2019) on multi-institutional data (accuracy of 64.3% for three-class OS prediction) using Convolutional Neural Networks, where CNN based methods are commonly expected to provide much-improved performance compared to traditional methods.…”

Ensemble learning of multiview CNN models for survival time prediction of brain tumor patients using multimodal MRI scans

“…As shown in Table 5 Chen [48] showed that extracted handcrafted features, such as intensity, shape, texture and wavelet, from manually delineated tumor regions in pre-surgical axial T1Ce modality, and subsequently combined with clinical data allowed stratifying patients' survival into a low-or high-risk group with an AUC of 0.851. In [12] using SVM classifier and 2D texture features extracted from slices with the largest tumor size that are manually segmented by two experienced radiologists, authors compared the performance of four MRI modalities when used individually and in combination for classifying survival into two groups. The result showed that when using only T1Ce and the four MRI modalities separately, both models achieved nearly equal accuracy and AUC value of 80.7%, and 0.79, respectively.…”

“…Therefore, combining the information properly from different modalities and across the three planes may enhance the diagnostic performance. However, when evaluating survival prediction of glioma patients using these multimodal images, which modalities and projections are the most effective, and whether combinations of multiple modalities can improve the performance are still unclear and not investigated very well [12]. MRI also comes at the cost of generating a large number of images per patient, exceeding the capacity of available physicians, particularly in middle-and low-income countries [13].…”

Deciphering the glioblastoma phenotype by computed tomography radiomics

“…gaussian and Z-score normalisation) and voxel reslicing [49][50][51]. So far, several studies have investigated normalized MRI images by means of radiomics and demonstrated cindices between 0.62 and 0.85 for multivariate models with either MR features alone or combined with clinical features such as age and performance status or molecular markers such as MGMT [52][53][54][55][56][57][58][59][60]. The combined (C + R)PS model presented in this study returned an apparent c-index of 0.69, which is comparable to the studies evaluating radiomics on MRI.…”