Despite advancements in molecular and histopathologic characterization of pediatric low-grade gliomas (pLGGs), there remains significant phenotypic heterogeneity among tumors with similar ...categorizations. We hypothesized that an unsupervised machine learning approach based on radiomic features may reveal distinct pLGG imaging subtypes.
Multi-parametric MR images (T1 pre- and post-contrast, T2, and T2 FLAIR) from 157 patients with pLGGs were collected and 881 quantitative radiomic features were extracted from tumorous region. Clustering was performed using K-means after applying principal component analysis (PCA) for feature dimensionality reduction. Molecular and demographic data was obtained from the PedCBioportal and compared between imaging subtypes.
K-means identified three distinct imaging-based subtypes. Subtypes differed in mutational frequencies of BRAF (p < 0.05) as well as the gene expression of BRAF (p<0.05). It was also found that age (p < 0.05), tumor location (p < 0.01), and tumor histology (p < 0.0001) differed significantly between the imaging subtypes.
In this exploratory work, it was found that clustering of pLGGs based on radiomic features identifies distinct, imaging-based subtypes that correlate with important molecular markers and demographic details. This finding supports the notion that incorporation of radiomic data could augment our ability to better characterize pLGGs.
Tumor-associated macrophages (TAMs) play an important role in tumor immunity and comprise of subsets that have distinct phenotype, function, and ontology. Transcriptomic analyses of human ...medulloblastoma, the most common malignant pediatric brain cancer, showed that medulloblastomas (MBs) with activated sonic hedgehog signaling (SHH-MB) have significantly more TAMs than other MB subtypes. Therefore, we examined MB-associated TAMs by single-cell RNA sequencing of autochthonous murine SHH-MB at steady state and under two distinct treatment modalities: molecular-targeted inhibitor and radiation. Our analyses reveal significant TAM heterogeneity, identify markers of ontologically distinct TAM subsets, and show the impact of brain microenvironment on the differentiation of tumor-infiltrating monocytes. TAM composition undergoes dramatic changes with treatment and differs significantly between molecular-targeted and radiation therapy. We identify an immunosuppressive monocyte-derived TAM subset that emerges with radiation therapy and demonstrate its role in regulating T cell and neutrophil infiltration in MB.
Display omitted
•Sonic Hedgehog (SHH) subgroup of medulloblastoma (MB) recruits diverse macrophages•Radiation or molecular-targeted therapy alters macrophage distribution in SHH-MB•Radiation recruits immunosuppressive monocyte-derived macrophages (TAMoMacs) in SHH-MB•Radiation-induced TAMoMacs regulate CD8 T cell and neutrophil numbers in SHH-MB
Dang et al. show that the sonic hedgehog subgroup of medulloblastoma (SHH-MB) contains macrophages derived from microglia and circulating monocytes. Radiation therapy, but not treatment targeting the SHH pathway, led to recruitment of immunosuppressive monocyte-derived macrophages that reduced T cells and neutrophils in the tumor microenvironment.
Medulloblastoma is a highly heterogeneous pediatric brain tumor with five molecular subtypes, Sonic Hedgehog TP53-mutant, Sonic Hedgehog TP53-wildtype, WNT, Group 3, and Group 4, defined by the World ...Health Organization. The current mechanism for classification into these molecular subtypes is through the use of immunostaining, methylation, and/or genetics. We surveyed the literature and identified a number of RNA-Seq and microarray datasets in order to develop, train, test, and validate a robust classifier to identify medulloblastoma molecular subtypes through the use of transcriptomic profiling data. We have developed a GPL-3 licensed R package and a Shiny Application to enable users to quickly and robustly classify medulloblastoma samples using transcriptomic data. The classifier utilizes a large composite microarray dataset (15 individual datasets), an individual microarray study, and an RNA-Seq dataset, using gene ratios instead of gene expression measures as features for the model. Discriminating features were identified using the limma R package and samples were classified using an unweighted mean of normalized scores. We utilized two training datasets and applied the classifier in 15 separate datasets. We observed a minimum accuracy of 85.71% in the smallest dataset and a maximum of 100% accuracy in four datasets with an overall median accuracy of 97.8% across the 15 datasets, with the majority of misclassification occurring between the heterogeneous Group 3 and Group 4 subtypes. We anticipate this medulloblastoma transcriptomic subtype classifier will be broadly applicable to the cancer research and clinical communities.
Abstract
PURPOSE
Although WHO grade I meningiomas are considered ‘benign’ tumors, an elevated Ki-67 is one crucial factor that has been shown to influence clinical outcomes. In this study, we use ...standard pre-operative MRI and develop a machine learning (ML) model to predict the Ki-67 in WHO grade I meningiomas.
METHODS
A retrospective analysis was performed of 306 patients that underwent surgical resection. The mean and median Ki-67 of tumor specimens were 4.84 ± 4.03% (range: 0.3–33.6) and 3.7% (Q1:2.3%, Q3:6%), respectively. Pre-operative MRI was used to perform radiomic feature extraction (N=2,520) followed by ML modeling using least absolute shrinkage and selection operator (LASSO) wrapped with support vector machine (SVM) through nested cross-validation on a discovery cohort (N=230), to stratify tumors based on Ki-67 < 5% and ≥ 5%. A replication cohort (N=76) was kept ‘unseen’ in order to provide insights regarding the generalizability of our predictive model.
RESULTS
A total of 60 radiomic features extracted from seven different MRI sequences were used in the final model. With this model, an AUC of 0.84 (95% CI: 0.78-0.90), with associated sensitivity and specificity of 84.1% and 73.3%, respectively, were achieved in the discovery cohort. The selected features in the trained predictive model were then applied to the subjects of the replication cohort and the model was applied independently in this cohort. An AUC of 0.83 (95% CI: 0.73-0.94), with a sensitivity of 82.6% and specificity of 85.5% was obtained for this independent testing. Furthermore, the model performed commendably when applied to all skull base and non-skull base tumors in our patient cohort, evidenced by comparable AUC values of 0.86 and 0.83, respectively.
CONCLUSION
The results of this study may provide enhanced diagnostics to the surgeon pre-operatively such that it can guide surgical strategy and individual patient treatment paradigms.
INTRODUCTION:
An elevated Ki-67 is one crucial factor that influences meningioma behavior. Machine learning(ML) using radiomic feature analysis can identify phenotypic pixel-level imaging signatures ...for enhanced tumor diagnostics.
METHODS:
Radiomic feature extraction using least-absolute-shrinkage-and-selection-operator(LASSO) wrapped with support vector machine (SVM) through nested cross-validation is used to stratify tumors based on Ki-67 <5% and ≥5%. Clinical outcomes were evaluated based on the predictive power of this algorithm.
RESULTS:
343 patients are included (WHO grade I:291, grade II:43, grade III:9). Overall mean follow-up time is 33.9 months (range: 0-105). The rate of recurrence is 14.4%, 44.2% and 77.8% for grade I, II, and III tumors, respectively. The mean Ki-67% for grade 1, 2 and 3 meningiomas is 4.79+/-3.87 (range: 0.3-33.6), 16.07+/-13.83 (range: 1.5-49), and 35.7+/-13.3 (range: 18-57.4), respectively (p = 0.03). However, there is no difference in tumor and peritumoral edema volumes between meningioma WHO grades 1-3. A total of 46 high-ranking radiomic features were selected to build a ML model. ROC curves for the ML algorithm reveals AUC’s of 0.83 95% CI: 0.78-0.89 and 0.84 0.75-0.94 for the discovery (N = 257) and validation (N = 86) cohort, respectively, for classifying Ki-67% based on a 5% cutoff, independent of WHO grade. Kaplan-Meier curves using the ML algorithm reveals decreased progression-free-survival (PFS) for predicted Ki-67 >5% (p < 0.001) for all tumors in the cohort, as well as in sub-analyses of non-skull base tumors (p < 0.001) and skull-base tumors (p < 0.001). A comparison of histopathological PFS versus ML-predicted outcomes based on Ki-67 reveals good concordance for grade I, II and III tumors.
CONCLUSIONS:
ML using radiomic feature analysis can be used to stratify meningiomas based on Ki-67 with excellent accuracy. The predictive power of this model reveals distinctly divergent PFS outcomes, and can be used to guide treatment strategy.
Abstract
Current response assessment in pediatric brain tumors (PBTs), as recommended by the Response Assessment in Pediatric Neuro-Oncology (RAPNO) working group, relies on 2D measurements of ...changes in tumor size. However, there is growing evidence of underestimation of tumor size in PBTs using 2D compared to volumetric (3D) measurement approach. Accordingly, automated methods that reduce manual burden and intra- and inter-rater variability in segmenting tumor subregions and volumetric evaluations are warranted to facilitate tumor response assessment of PBTs. We have developed a fully automatic deep learning (DL) model using the nnUNet architecture on a large cohort of multi-institutional and multi-histology PBTs. The model was trained on widely available standard multiparametric MRI sequences (T1-pre, T1-post, T2, T2-FLAIR) for segmentation of the whole tumor and RAPNO-recommended subregions, including enhancing tumor (ET), non-enhancing tumor (NET), cystic components (CC), and peritumoral edema (ED). As a prerequisite step for accurate tumor segmentation, we also generated another DL model based on DeepMedic for brain extraction from mpMRIs. The models were trained on an institutional cohort of 151 subjects and independently tested on 64 subjects from the internal and 29 patients from external institutions. The trained models showed excellent performance with median Dice scores of 0.98±0.02/0.97±0.02 for brain tissue segmentation, 0.92±0.08/0.90±0.17 for whole tumor segmentation, 0.76±0.31/0.87±0.29 for ET subregion, and 0.82±0.15/0.80±0.28 for segmentation of non-enhancing components (combination of NET, CC, and ED) in internal/external test sets, respectively. The automated segmentation demonstrated strong agreement with expert segmentations in volumetric measurement of tumor components, with Pearson’s correlation coefficients of 0.97, 0.97, 0.99, and 0.79 (p<0.0001) for ET, NET, CC, and ED regions, respectively. Our proposed multi-institutional and multi-histology automated segmentation method has the potential to aid clinical neuro-oncology practice by providing reliable and reproducible volumetric measurements for treatment response assessment.
Abstract
Pediatric low-grade glioma (pLGG) encompasses a variety of tumor subtypes with heterogeneous treatment response and relatively long progression-free survival (PFS). Radiomics may serve as a ...non-invasive and in-vivo tool for early prediction of PFS as a surrogate marker for treatment response and to objectively gauge the efficacy of novel treatment strategies. Here, we present a multivariate model based on radiomic features and clinical variables for risk stratification of pLGGs in terms of PFS and seek associations of the predicted risk groups and mutations in key molecular markers using data from PedCBioportal. Pre-operative multi-parametric MRI scans (T1-pre, T1-post, T2, T2-FLAIR) of 129 patients with newly diagnosed pLGG (median age, 7.76, range, 0.35-19.58 years; median PFS, 28.5, range, 1.1-124.8 months) were collected and quantitative radiomic features (n = 881) were extracted. A multivariate Cox proportional hazard’s (Cox-PH) regression model was fitted based on clinical (age, sex, and extent of tumor resection) and radiomic variables using 4-fold cross-validation. A subset of radiomic features (n = 27) that were most predictive of PFS was selected by applying Elastic Net regularization penalty during Cox-PH model fitting. High-, medium- and low-risk groups were determined based on model predictions. Cox-PH modeling showed excellent performance for prediction of PFS as suggested by the concordance index of 0.78. Radiogenomic assessment (data available in 94/129 patients) showed more enrichment of mutations in NF1 and RB1 genes in the high-risk group, as compared to the low- and medium-risk groups. We showed the potential value of radiomics in providing upfront prediction of PFS, which may further be used as an added treatment arm for early assessment of treatment response of the pLGG patients enrolled in the clinical trials. In the next step of this work, we will expand the cohort and cross-validate these results in an external cohort.
Abstract
Recent studies have shown preliminary evidence for differentiation of the tumor microenvironment (TME) and immune landscape between molecularly-defined medulloblastoma (MB) subtypes. ...Identifying radiological correlates of these TME patterns could establish a non-invasive method of immune profile characterization for guiding patient-centered therapies. Here, we examine immune profiles between MB subtypes using data from Open Pediatric Brain Tumor Atlas (OpenPBTA), and their relationship to tumor measurements from pre-operative MRIs. We identified a retrospective cohort of 94 pediatric MB patients with available molecular subtyping and immune profiles (36 cell types) from bulk gene expression data. A random forest analysis was used to classify the four MB subtypes based on immune profiles. Four cell types had high impact on classification performance: plasmacytoid dendritic cells (PDC; 25.8% accuracy decrease when randomized), hematopoietic stem cells (HSC; 21.9%), plasma B cells (20.3%), and cancer associated fibroblasts (18.8%). Pairwise comparisons revealed SHH and WNT tumors had significantly higher numbers of fibroblasts and HSCs compared to Group3/Group4. We also found novel evidence for significantly lower amounts of plasma B cells in the SHH group, and high PDC levels in Group4, followed by Group3, and low PDC in SHH/WNT. Multi-parametric MRI scans for 39 patients were used to segment tumor volumes. Overall tumor volume was significantly correlated with composite stroma scores (R = 0.34, p = 0.036). Additionally, patients with higher volumes of gadolinium contrast-enhancing compared to non-enhancing components had higher immune (R = 0.42, p = 0.009) and microenvironment (summed immune and stromal cell types; R = 0.44, p = 0.006) scores, regardless of their molecular subtype. Together, our results demonstrate: (1) the use of rich immune profiles for differentiating molecular subtypes of MB and their unique TME characterization; and (2) initial evidence for radiological correlates of these profiles based on pre-operative imaging collected through standard practices.
Abstract
Volumetric measurements of whole tumor and its components on MRI scans, facilitated by automatic segmentation tools, are essential to reduce inter-observer variability in monitoring tumor ...progression and response assessment for pediatric brain tumors. Here, we present a fully automatic segmentation model based on deep learning that reliably delineates the tumor components recommended by the Response Assessment in Pediatric Neuro-Oncology (RAPNO) working group for evaluation of treatment response. Multi-parametric MRI (mpMRI) scans (T1-pre, T1-post, T2, and T2-FLAIR), acquired on multiple MRI scanners with different field strengths and vendors, for a cohort of 218 pediatric patients with a variety of histologically confirmed brain tumor subtypes were collected. The mpMRI scans were co-registered and manually segmented by experienced neuroradiologists in consensus to identify the tumor subregions including the enhancing tumor (ET), non-enhancing tumor (NET), cystic components (CC), and peritumoral edema (ED) regions. A convolutional neural network model based on DeepMedic architecture was trained using mpMRI scans as the inputs for segmentation of the whole tumor and subregions. The trained model showed excellent performance in segmentation of the whole tumor, as suggested by median dice of 0.90/0.85 for validation (n = 44)/independent test (n = 22) sets. ET and non-enhancing components (union of NET, CC, and ED) were segmented with median dice scores of 0.78/0.84 and 0.76/0.74 for validation/test sets, respectively. The automated and manual segmentations demonstrated strong agreement in estimating VASARI (Visually AcceSAble Rembrandt Images) MRI features with Pearson’s correlation coefficient R > 0.75 (p < 0.0001) for ET, NET, CC, and ED components. Our proposed automated segmentation method developed based on MRI scans acquired with different protocols, equipment, and from a variety of brain tumor subtypes, shows potential application for reliable and generalizable volumetric measurements which can be used for treatment response assessment in clinical trials.