Search

The ability to precisely label and visualize tumors in tissues can improve the accuracy of surgical resections compared to standard of care, which relies on visual inspection and palpation. While fluorescence intensity-based imaging is being evaluated for surgical guidance, variable tumor uptake and incomplete clearance of fluorescent probes reduces tumor vs. normal classification accuracy. Here we demonstrate that the fluorescence lifetime (FLT) of multiple types of solid tumors is longer than the FLT of healthy tissues in patients systemically injected with indocyanine green (ICG), an FDA approved near infrared dye. We show that this cancer-specific lifetime shift can distinguish tumor from normal tissue both at a cellular level using microscopy and in large specimens using wide-field imaging, with an accuracy of over 97% across multiple patients. Unlike intensity, which is a system-specific parameter that depends on tumor dye uptake and depth in tissue, FLT is a system-independent photophysical property that can be quantified in thick tissues. Our study suggests that FLT imaging with ICG can be immediately used to improve the accuracy of cancer surgeries.

Extraneural metastases of glioblastoma (GBM), although rare, are becoming an increasingly recognized occurrence. Currently, the biological mechanism underlying this rare occurrence is not understood. To explore the potential genomic drivers of extraneural metastasis in GBM, we present the molecular features of 4 extraneural metastatic GBMs, along with a comprehensive review and analysis of previously reported cases that had available molecular characterization. In addition to our 4 cases, 42 patients from 35 publications are reviewed. To compare the molecular profiles between GBM cases with extraneural metastasis and the general GBM population, genomic data from GBM samples in The Cancer Genome Atlas (TCGA) database were also analyzed. We found that 64.5% (20/31) of the cases with extraneural metastasis that were tested for TP53 changes had at least 1 TP53 pathogenic variant detected in either 1 or both primary and metastatic tumors. In contrast, TP53 mutation was significantly less frequent in the unselected GBM from TCGA (22.6%, 56/248) (P=0.000). In addition, O-6-methylguanine-DNA methyltransferase (MGMT) promoter methylation was more common in unselected TCGA GBM cases (48.6%, 170/350) than in cases with extraneural metastasis (31.8%, 7/22), although not statistically significant. Although isocitrate dehydrogenase (IDH) mutation is a rare occurrence in high-grade astrocytomas, IDH-mutant grade 4 astrocytomas are at least as likely to metastasize as IDH wild-type GBMs; 3 metastatic cases definitively harbored an IDH1 (p.R132H) mutation in our analysis. Our findings not only provide potential biomarkers for earlier screening of extraneural metastasis, but could also suggest clues to understanding biological mechanisms underlying GBM metastasis, and for the development of therapeutic modalities.

PURPOSE T2-FLAIR mismatch (T2FM) is a highly specific imaging biomarker for isocitrate dehydrogenase (IDH) mutation in low-grade gliomas. Previous T2FM studies are inconsistent for glioblastoma (GBM)/grade-4 glioma, partly due to low IDH-mutation prevalence in high-grade gliomas. We leveraged a large multi-institutional GBM/grade-4 glioma cohort to analyze the association of partial T2FM and IDH-mutation (T2-hyperintense, FLAIR-hypointense, nonenhancing, nonedema). METHODS We analyzed preoperative MRI of 1500 pathologically confirmed GBM/grade-4 gliomas with known IDH-mutation status from the ReSPOND consortium, consisting of the following institutions (sample size): Ivy GBM Atlas Project (33), Catalan Institute of Oncology (132), Case Western Reserve University/University Hospitals (132), New York University (55), Ohio State University (25), University of Pennsylvania (641), University Hospital Río Hortega (16), Yonsei University Health System (118), The Cancer Imaging Archive (93), Thomas Jefferson University (48), Tata Memorial Hospital (22), University of Pittsburgh Medical Center (156), and Washington University School of Medicine in St. Louis (57). Sequences were co-registered to a common anatomic atlas. Continuous variables were compared by t-test and categorical variables by Χ 2-test. RESULTS 71 (4.7%) were IDH-mutants, significantly younger (43±1 v. 62±12 years, p=5x10-37), and more likely to exhibit partial T2FM (20% v. 0.4%, p=1x10-43), frontal lobe predominance (68% v. 29%, p=7x10-12), nonenhancing components (T2/FLAIR-intermediate signal, nonedema; 45% v. 9%, p=1x10-22), and cystic components (smooth margins, no/minimal enhancement, homogeneous FLAIR suppression; 17% v. 3%, p=7x10-11) than IDH-wildtypes. 20 cases had partial T2FM (14 IDH-mutant, 6 IDH-wildtype). Sensitivity of partial T2FM for IDH-mutation was 19.7%, specificity 99.6%, positive predictive value 70%, and negative predictive value 96.1%. Subset analysis of 983 IDH-wildtypes with known MGMT methylation status (406 MGMT-hypermethylated) showed frontal lobe predominance was more common in MGMT-hypermethylated than MGMT-unmethylated (39.4% v. 24.3%, p=.02); other imaging characteristics did not significantly differ. CONCLUSIONS Partial T2FM is a highly specific imaging biomarker for IDH-mutation in GBM/grade-4 glioma.

OBJECTIVES Although the impact of surgery in glioblastoma (GBM) is well-established, there is insufficient data to determine how surgical resection affects overall survival (OS). The impact of other GBM genomic alterations on EOR-induced survival remains poorly understood. Here, we leverage a large cohort of GBM patients with genomic data to establish the impact of genomic alterations on resection-dependent overall survival in GBM. METHODS 537 IDH WT GBMs from UPMC were analyzed using Glioseq next generation sequencing panel. Using molecular profiles derived from Glioseq data, semi-supervised Monte Carlo reference-based consensus clustering was used to define 5 distinct groups. Cox Proportional Hazards models were used to identify predictors of OS within molecular groups. RESULTS Clustering of 537 IDH wild-type GBM identified 5 distinct groups: 1: TERT mutated/MGMT methylated/EGFR mutated or gain; 2: TERT mutated /MGMT methylated/p53 mutated; 3: TERT WT; 4: TERT mutated/MGMT methylated/CDKN2A loss; 5: TERT mutated/MGMT unmethylated. Thresholds in postoperative residual contrast enhancing volume (rCEV) and residual non-contrast enhancing volume (rNCEV) were identified which marked significant differences in median overall survival (MOS) between molecular groups in rCEV -Group 1: 11.7cc, Group 2: 14.4cc, Group 3: 1.1cc, Group 4: 0.1cc, and Group 5 1.1cc. Sex, age, KPS >70, temozolomide, and radiation along with thresholded rCEV and rNCEV were used in multivariate Cox analyses. Groups demonstrated distinct predictors of OS: radiation only in Group 1 tumors, low rCEV in Group 2 tumors, younger age, temodar, low CEV, and KPS >70 in Group 3 tumors, female sex, younger age, radiation, KPS >70 low rCEV in Group 4 tumors, and KPS >70, temodar, and low rCEV in Group 5 tumors. CONCLUSIONS Clustering results suggest that there are clinically distinct molecular groups of IDH wild-type glioblastomas. Age, sex, KPS >70, temozolomide treatment, radiation therapy, and EOR have varying affects on these molecular groups.

PURPOSE Glioblastoma is extremely infiltrative with malignant cells extending beyond the enhancing rim where recurrence inevitably occurs, despite aggressive multimodal therapy. We hypothesize that important characteristics of peritumoral tissue heterogeneity captured and analyzed by multi-parametric MRI and artificial intelligence (AI) methods are generalizable in the updated multi-institutional ReSPOND (Radiomics Signatures for PrecisiON Diagnostics) consortium and predictive of neoplastic infiltration and future recurrence. METHODS We used the most recent update of the ReSPOND consortium to evaluate and further refine generalizability of our methods with different scanners and acquisition settings. 179 de novo glioblastoma patients with available T1, T1Gd, T2, T2-FLAIR, and ADC sequences at pre-resection baseline and after complete resection with subsequent pathology-confirmed recurrence were included. To establish generalizability of the predictive models, training and testing of the refined AI model was performed through Leave-One-Institution-Out-Cross-Validation schema. The multi-institutional cohort consisted of the Hospital of the University of Pennsylvania (UPenn, 124), Case Western Reserve University/University Hospitals (CWRU/UH, 27), New York University (NYU, 13), Ohio State University (OSU, 13), and University Hospital Río Hortega (RH, 2). Features extracted from pre-resection MRI were used to build the model predicting the spatial pattern of subsequent tumor recurrence. These predictions were evaluated against regions of pathology-confirmed post-resection recurrence. RESULTS Our model predicted the locations that later harbored tumor recurrence with overall odds ratio (99% CI)/AUC (99% CI), 12.0(11.8-12.2)/0.80(0.76-0.85), and per institute, CWRU/UH, 11.0(10.7-11.3)/0.80 (0.64-0.97); NYU, 7.0(6.7-7.3)/0.78(0.56-1.00); OSU, 18.3(17.5-19.1)/0.83(0.54-1.00); RH, 40.0(35.3-45.5)/0.93(0.00-1.00); UPenn, 8.00(7.7-8.3)/0.80(0.75-0.84). CONCLUSION This study provides extensive multi-institutional validated evidence that machine learning tools can identify peritumoral neoplastic infiltration and predict location of future recurrence, by decrypting the MRI signal heterogeneity in peritumoral tissue. Our analyses leveraged the unique dataset of the ReSPOND consortium, which aims to develop and validate AI-based biomarkers for individualized prediction and prognostication and establish generalizability in a multi-institutional setting.

Background Gliomas represent a biologically heterogeneous group of primary brain tumors with uncontrolled cellular proliferation and diffuse infiltration that renders them almost incurable, thereby leading to a grim prognosis. Recent comprehensive genomic profiling has greatly elucidated the molecular hallmarks of gliomas, including the mutations in isocitrate dehydrogenase 1 and 2 (IDH1 and IDH2), loss of chromosomes 1p and 19q (1p/19q), and epidermal growth factor receptor variant III (EGFRvIII). Detection of these molecular alterations is based on ex vivo analysis of surgically resected tissue specimen that sometimes is not adequate for testing and/or does not capture the spatial tumor heterogeneity of the neoplasm. Methods We developed a method for noninvasive detection of radiogenomic markers of IDH both in lower-grade gliomas (WHO grade II and III tumors) and glioblastoma (WHO grade IV), 1p/19q in IDH-mutant lower-grade gliomas, and EGFRvIII in glioblastoma. Preoperative MRIs of 473 glioma patients from 3 of the studies participating in the ReSPOND consortium (collection I: Hospital of the University of Pennsylvania [HUP: n = 248], collection II: The Cancer Imaging Archive [TCIA; n = 192], and collection III: Ohio Brain Tumor Study [OBTS, n = 33]) were collected. Neuro-Cancer Imaging Phenomics Toolkit (neuro-CaPTk), a modular platform available for cancer imaging analytics and machine learning, was leveraged to extract histogram, shape, anatomical, and texture features from delineated tumor subregions and to integrate these features using support vector machine to generate models predictive of IDH, 1p/19q, and EGFRvIII. The models were validated using 3 configurations: (1) 70–30% training–testing splits or 10-fold cross-validation within individual collections, (2) 70–30% training–testing splits within merged collections, and (3) training on one collection and testing on another. Results These models achieved a classification accuracy of 86.74% (HUP), 85.45% (TCIA), and 75.15% (TCIA) in identifying EGFRvIII, IDH, and 1p/19q, respectively, in configuration I. The model, when applied on combined data in configuration II, yielded a classification success rate of 82.50% in predicting IDH mutation (HUP + TCIA + OBTS). The model when trained on TCIA dataset yielded classification accuracy of 84.88% in predicting IDH in HUP dataset. Conclusions Using machine learning algorithms, high accuracy was achieved in the prediction of IDH, 1p/19q, and EGFRvIII mutation. Neuro-CaPTk encompasses all the pipelines required to replicate these analyses in multi-institutional settings and could also be used for other radio(geno)mic analyses.

PURPOSE To construct a multi-institutional radiomic model that supports upfront prediction of progression-free survival (PFS) and recurrence pattern (RP) in patients diagnosed with glioblastoma multiforme (GBM) at the time of initial diagnosis. PATIENTS AND METHODS We retrospectively identified data for patients with newly diagnosed GBM from two institutions (institution 1, n = 65; institution 2, n = 15) who underwent gross total resection followed by standard adjuvant chemoradiation therapy, with pathologically confirmed recurrence, sufficient follow-up magnetic resonance imaging (MRI) scans to reliably determine PFS, and available presurgical multiparametric MRI (MP-MRI). The advanced software suite Cancer Imaging Phenomics Toolkit (CaPTk) was leveraged to analyze standard clinical brain MP-MRI scans. A rich set of imaging features was extracted from the MP-MRI scans acquired before the initial resection and was integrated into two distinct imaging signatures for predicting mean shorter or longer PFS and near or distant RP. The predictive signatures for PFS and RP were evaluated on the basis of different classification schemes: single-institutional analysis, multi-institutional analysis with random partitioning of the data into discovery and replication cohorts, and multi-institutional assessment with data from institution 1 as the discovery cohort and data from institution 2 as the replication cohort. RESULTS These predictors achieved cross-validated classification performance (ie, area under the receiver operating characteristic curve) of 0.88 (single-institution analysis) and 0.82 to 0.83 (multi-institution analysis) for prediction of PFS and 0.88 (single-institution analysis) and 0.56 to 0.71 (multi-institution analysis) for prediction of RP. CONCLUSION Imaging signatures of presurgical MP-MRI scans reveal relatively high predictability of time and location of GBM recurrence, subject to the patients receiving standard first-line chemoradiation therapy. Through its graphical user interface, CaPTk offers easy accessibility to advanced computational algorithms for deriving imaging signatures predictive of clinical outcome and could similarly be used for a variety of radiomic and radiogenomic analyses.

Fusions involving neurotrophic tyrosine receptor kinase (NTRK) genes are detected in ≤2% of gliomas and can promote gliomagenesis. The remarkable therapeutic efficacy of TRK inhibitors, which are among the first Food and Drug Administration-approved targeted therapies for NTRK-fused gliomas, has generated significant clinical interest in characterizing these tumors. In this multi-institutional retrospective study of 42 gliomas with NTRK fusions, next generation DNA sequencing (n = 41), next generation RNA sequencing (n = 1), RNA-sequencing fusion panel (n = 16), methylation profile analysis (n = 18), and histologic evaluation (n = 42) were performed. All infantile NTRK-fused gliomas (n = 7) had high-grade histology and, with one exception, no other significant genetic alterations. Pediatric NTRK-fused gliomas (n = 13) typically involved NTRK2, ranged from low- to high-histologic grade, and demonstrated histologic overlap with desmoplastic infantile ganglioglioma, pilocytic astrocytoma, ganglioglioma, and glioblastoma, among other entities, but they rarely matched with high confidence to known methylation class families or with each other; alterations involving ATRX, PTEN, and CDKN2A/2B were present in a subset of cases. Adult NTRK-fused gliomas (n = 22) typically involved NTRK1 and had predominantly high-grade histology; genetic alterations involving IDH1, ATRX, TP53, PTEN, TERT promoter, RB1, CDKN2A/2B, NF1, and polysomy 7 were common. Unsupervised principal component analysis of methylation profiles demonstrated no obvious grouping by histologic grade, NTRK gene involved, or age group. KEGG pathway analysis detected methylation differences in genes involved in PI3K/AKT, MAPK, and other pathways. In summary, the study highlights the clinical, histologic, and molecular heterogeneity of NTRK-fused gliomas, particularly when stratified by age group.

Differentiation of true tumor progression (TP) from pseudoprogression (PsP) is a major unmet need in patients with glioblastoma. [18F]Fluciclovine is a synthetic amino acid PET radiotracer that is FDA-approved in biochemical recurrence in prostate cancer. The study aim was to assess the value of [18F]Fluciclovine PET in differentiation of histologically confirmed (“true”) TP and PsP in post-treatment of glioblastoma. METHODS: 23 patients with glioblastoma with new contrast-enhancing lesions or lesions showing increased enhancement ( > 25% increase) on standard MRI after completion of radiation underwent 60-minutes dynamic [18F]Fluciclovine PET imaging. Patients subsequently underwent resection of enhancing lesion and tumor percentage vs. treatment-related changes were quantified on histopathology. Patients were considered "true” TP if tumor represented ≥ 50% of the resected specimen, mixed TP-PsP if < 50% and > 10%, and PsP if tumor represented ≤ 10%. Summed 30- to 40-minute post-injection PET images were used to measure SUVpeak and SUVmax (g/mL units). RESULTS: 15 patients with “true” TP, 3 with mixed TP-PsP, and 5 with PsP were included. There was a positive correlation between SUVpeak by PET and tumor percentage by histology (Rho= 0.56, p= 0.006). Patients who demonstrated “true” TP had significantly higher SUVpeak compared to patients with histological PsP (4.8±1.6 vs 2.9± 1.0, p= 0.02, AUC= 0.91, n=20). SUVpeak cut-off of 3.3 provided 93% sensitivity, 80% specificity, and 90% accuracy for differentiation of “true” TP from PsP. Patients with “true” TP/mixed TP-PsP also had significantly higher SUVpeak than patients with PsP (4.6±1.5 vs 2.9± 1.0, p= 0.03, AUC= 0.88, n=23). SUVmax and partial volume-corrected SUVpeak and SUVmax exhibited similar performance. CONCLUSION: Our results indicated that [18F]Fluciclovine PET imaging can accurately differentiate “true” TP from PsP. Further studies are required to confirm these promising early results and determine optimal criteria for interpreting [18F]Fluciclovine PET to distinguish PsP from TP.

BACKGROUND Tumor-associated macrophages (TAMs) are a key component of glioblastoma (GBM) tumor microenvironment. Considering the differential role of different TAM phenotypes in iron metabolism with the M1 phenotype storing intracellular iron, and M2 phenotype releasing iron in the tumor microenvironment, here we investigated non-invasive quantitative susceptibility mapping (QSM) and T2* MRI relaxometry to quantify iron as imaging biomarkers for TAMs in adult patients with GBM. METHODS In this prospective study, 21 adult patients with GBM were enrolled between 2016 to 2019. Patients underwent a 3D single echo gradient echo sequence in addition to standard anatomical sequences on a 3 Tesla MRI. QSM images were reconstructed using the morphology-enabled dipole inversion (MEDI) algorithm. In 3 subjects, ex vivo imaging of surgical specimens was performed on a Bruker 9.4 Tesla 8.9 cm vertical bore MR using 3D multi-echo GRE scans, and R2* (1/T2*) maps were generated by a pixel-wise monoexponential fitting. Each specimen was stained with hematoxylin and eosin, as well as with CD68, CD86, CD206, and L-Ferritin. RESULTS Significant positive correlation was observed between mean susceptibility for the tumor enhancing zone and the L-ferritin positivity percent (r =0.56, p=0.018) and the combination of tumor’s enhancing zone and necrotic core and the L-Ferritin positivity percent (r=0.72; p=0.001). Moreover, the mean susceptibility significantly correlated with positivity percent for CD68 (ρ=0.52, p=0.034) and CD86 (r=0.7 p=0.001), but not for CD206 (ρ=0.09; p=0.7). There was a positive correlation between mean R2* values and CD68 positive cell counts (r =0.6, p=0.016). Similarly, mean R2* values significantly correlated with CD86 (r=0.54, p=0.03) but not with CD206 (r=0.15, p=0.5). High mean susceptibility in the necrotic core was associated with inferior PFS (hazard ratio,4.5, p=0.016). CONCLUSION MR Susceptibility Imaging can quantify the iron content of GBM and provide a non-invasive method for TAM quantification and phenotyping.

Purpose Tumor-associated macrophages (TAMs) are a key component of glioblastoma (GBM) microenvironment. Considering the differential role of different TAM phenotypes in iron metabolism with the M1 phenotype storing intracellular iron, and M2 phenotype releasing iron in the tumor microenvironment, we investigated MRI to quantify iron as an imaging biomarker for TAMs in GBM patients. Methods 21 adult patients with GBM underwent a 3D single echo gradient echo MRI sequence and quantitative susceptibility maps were generated. In 3 subjects, ex vivo imaging of surgical specimens was performed on a 9.4 Tesla MRI using 3D multi-echo GRE scans, and R2* (1/T2*) maps were generated. Each specimen was stained with hematoxylin and eosin, as well as CD68, CD86, CD206, and L-Ferritin. Results Significant positive correlation was observed between mean susceptibility for the tumor enhancing zone and the L-ferritin positivity percent (r =0.56, p=0.018) and the combination of tumor’s enhancing zone and necrotic core and the L-Ferritin positivity percent (r=0.72; p=0.001). The mean susceptibility significantly correlated with positivity percent for CD68 (ρ = 0.52, p=0.034) and CD86 (r=0.7 p=0.001), but not for CD206 (ρ = 0.09; p=0.7). There was a positive correlation between mean R2* values and CD68 positive cell counts (r =0.6, p=0.016). Similarly, mean R2* values significantly correlated with CD86 (r=0.54, p=0.03) but not with CD206 (r=0.15, p=0.5). Conclusion MR quantitative susceptibility mapping can quantify the iron content of GBM and provide a non-invasive method for TAM quantification and phenotyping.

PURPOSE The significant heterogeneity of glioblastoma is typically displayed on both phenotypical and molecular levels. Non-invasive in vivo approaches to characterize this heterogeneity would potentially facilitate personalized therapies. Here we leverage advanced unsupervised machine learning to integrate radiomic imaging features and genomics to identify distinct subtypes of glioblastoma. METHODS A retrospective cohort of 571 IDH-wildtype glioblastoma patients were collected with pre-operative multi-parametric MRI (T1, T1CE, T2, T2-FLAIR, DSC, DTI) scans (available in 462/571 patients) and targeted next-generation sequencing (NGS) data (available in 355/571 patients). Radiomic features (n= 971) were extracted from these MRI scans and a subset of 12 features were selected by L21-norm minimization. A total of 14 key driver genes in the 5 main pathways that are most frequently altered in glioblastoma were chosen. Subtypes were identified by a joint learning approach called Anchor-based Partial Multi-modal Clustering (APMC) on both radiomic and genomic modalities. RESULTS Three distinct glioblastoma subtypes were discovered by APMC based on 14-dimension NGS data together with 12 selected radiomic features representing characteristics from histograms, shape, and volumetric measures for different tumor sub-regions. The identified subtypes were 1) high-risk; 2) medium-risk; and 3) low-risk, in terms of their overall survival outcome in Kaplan-Meier analysis (p= 5.52e-6, log-rank test; HR= 1.51, 95%CI:1.20-1.74, Cox proportional hazard model). The three subtypes also displayed different molecular characteristics: subtype 1 exhibited increased frequency of mutation in [EGFR, PIK3CA, PTEN, NF1], subtype 3 showed frequently mutated [PDGFRA, ATRX], while subtype 2 did not show significant differences for mutations in any of these genes. CONCLUSION Our results revealed the synergistic value of integrated radiomic signatures and molecular characteristics for glioblastoma subtyping. Joint learning on both modalities could help better understand the molecular basis of phenotypical signatures of glioblastoma and further provide insights into the biologic underpinnings of tumor formation and progression.

BACKGROUND Glioblastoma is the most common and aggressive malignant adult brain tumor, with median overall survival (OS) of 14 months. Despite optimizations in the treatment regime and technological advancements, patient OS has remained unchanged for the last 20 years. Here we seek the prognostic stratification of glioblastoma patients, by virtue of distinct morphological patterns apparent in histopathology H&E-stained whole slide images (WSI) alone. METHODS We selected 271 (ntraining/nhold-out-test= 216/55) frozen digitized tissue sections, at 10x magnification, of untreated primary glioblastoma (IDH-wildtype) adult cases from the TCGA-GBM collection, without history of previous glioma and with post-operative OS > 30 days. The median OS across cases was used to divide them into short- (< 324 days) and long- ( > 390 days) survivors, after excluding 5 percentiles of cases around this median. Each WSI was partitioned in non-overlapping 256x256 patches. A pre-trained VGG16 AI model automatically extracted a 512-dimensional feature vector from each patch. Dimensionality was reduced, from 512 to 8, using principal component analysis, and distinct groups of morphological patterns were identified with unsupervised clustering. The number of distinct groups was decided based on the silhouette coefficient, assessing goodness of clustering. The proportions of these patterns within each WSI, describing the tumor’s spatial heterogeneity, were used to distinguish between short- and long-survivors using gradient boosting. RESULTS We identified morphological patterns of 5 distinct groups, descriptive of cellularity differences as well as artifactual content. Quantification of these patterns’ proportions resulted in a classification accuracy of 72.72% (short- vs long-survivors). CONCLUSIONS We identified morphologic patterns within a WSI, associated with shorter and longer OS. These may allow the clinical neuropathologist to i) provide additional prognostic information gleaned during microscopic assessment to the treating team, and ii) suggest avenues for further biological investigations towards contributing to our mechanistic understanding and potentially treating of glioblastoma.

PURPOSE Differentiation of tumor progression (TP) from pseudoprogression (PsP) is a major unmet need in post-treatment glioblastoma (GBM). 18F-Fluciclovine is a synthetic amino acid PET radiotracer with higher uptake in tumor tissue vs. areas of treatment-related change. We investigated the value of 18F-Fluciclovine PET for differentiating PsP from TP independent from and in combination with multi-parametric MRI. METHODS We prospectively enrolled 30 patients with GBM with a new or enlarging contrast-enhancing lesion on MRI after chemoradiotherapy who were planned for surgical resection of the lesion. Patients underwent pre-operative 18F-Fluciclovine PET and multi-parametric MRI. Following surgery, the relative percentages of viable tumor and therapy-related changes observed in histopathology were quantified. Patients were categorized as TP if viable tumor represented ≥ 50% of the specimen, mixed TP if < 50% and > 10%, and PsP if ≤ 10%. RESULTS 18 patients had TP, 4 had mixed TP, and 8 PsP. Patients with TP/mixed TP had a significantly higher 40-50 minutes SUVmax (6.64 + 1.88 vs 4.11± 1.52, p=0.009) and an SUVmax cut-off of 4.66 provided 90% sensitivity and 83% specificity for differentiation of TP/mixed TP from PsP (AUC=0.856). A maximum cerebral blood volume (CBVmax) cut-off of 3.67 provided 90% sensitivity and 71% specificity for differentiation of TP/mixed TP from PsP (AUC=0.779). Combining a 40-50 minutes SUVmax cut-off of 4.662 and a relative CBVmax cut-off of 3.67 provided 100% sensitivity and 80% specificity for differentiating TP/mixed TP from PsP (AUC=0.95). The time activity curve patterns and time to peaks were not different between the groups. Normalization of PET parameters to normal brain parenchyma were not helpful to differentiate the groups due to variability in radiotracer uptake in normal brain between subjects. CONCLUSION 18F-Fluciclovine PET uptake can accurately differentiate PsP from TP in GBM patients, with even more accurate differentiation achieved when combined with MRI.

PURPOSE There is evidence that molecular heterogeneity of glioblastoma is associated with heterogeneity of MR imaging signatures. Modern machine learning models, such as deep neural networks, provide a tool for capturing such complex relationships in high-dimensional datasets. This study leverages recent advances in visualizing neural networks to construct a radiogenomics coordinate system whose axes reflect the expression of imaging signatures of genetic mutations commonly found in glioblastoma. METHODS Multi-parametric MRI (mpMRI) scans (T1, T1-Gd, T2, T2-FLAIR, DSC, DTI) of 254 subjects with glioblastoma were retrospectively collected. Radiomics features, including histograms, morphologic and textural descriptors, were derived. Genetic markers were obtained through next generation sequencing (NGS) panel. A multi-label classification deep neural network was trained for predicting mutation status in key driver genes, EGFR, PTEN, NF1, TP53 and RB1. We utilized a nonlinear manifold learning method called Intensive Principal Component Analysis (InPCA), to visualize the output probability distributions from the trained model. The first three principal components (PCs) were selected for constructing the coordinate system. RESULTS The axes derived from InPCA analysis were associated with molecular pathways known to be implicated in glioblastoma: (1) Increasing values of PC1 were associated with primary involvement of P53 then RB1 then MAPK then RTK/PI3K; (2) Increasing values of PC2 were associated with primary involvement of RTK then RB1/P53/MAPK then PI3K; (3) Increasing values of PC3 were associated with primary involvement of MAPK then RB1/P53/RTK/PI3K. Imaging features significantly associated with each of three PCs (p< 0.05) were identified by Pearson correlation analysis. CONCLUSION Deep learning followed by nonlinear manifold embedding identifies a radiogenomics coordinate system spanned by three components which were associated with different molecular pathways of glioblastoma. The heterogeneity of radiogenomic signatures captured by this coordinate system offers in vivo biomarkers of the molecular heterogeneity of glioblastoma.

PURPOSE Glioblastoma, IDH-wildtype, is the most common primary malignant adult brain tumor with median overall survival (OS) of ~14 months, with little improvement over the last 20 years. We hypothesize that AI-based integration of quantitative tumor characteristics, independent of acquisition protocol and equipment, can reveal accurate generalizable prognostic stratification. We seek an AI-based OS predictor using routine clinically acquired MRI sequences, quantitatively evaluated across institutions of the ReSPOND (Radiomics Signatures for PrecisiON Diagnostics) consortium. METHODS We identified a retrospective cohort of 2,293 diffuse glioma (IDH-wildtype/-NOS/-NEC) patients from 22 geographically distinct institutions across 3 continents, with preoperative structural MRI scans. The entire tumor burden was automatically segmented into 3 sub-compartments, i.e., enhancing, necrotic, peritumoral T2-FLAIR abnormality. We developed our AI predictor by multivariate integration of i)patient age, ii)tumor sub-compartment volume normalized to brain volume, iii)spatial distribution characteristics (tumor location, distance to the ventricles, and laterality), and iv)morphologic descriptors (major axes’ length, axes’ ratio, extent, and number of tumors). The AI predictor returns a continuous value between 0-1, defining short-, intermediate-, and long-survivors based on thresholds on the 25th and 75th percentiles. Leave-One-Site-Out-Cross-Validation was used to assess the generalizability of our stratification. Kaplan-Meier survival curves were computed for OS analysis and evaluated by a Cox proportional hazards model for statistical significance and hazard ratios. RESULTS Survival analysis yielded a hazard ratio of 2.07 (95%CI, 2.06-2.08, p-value= 4.8e-102) for patient stratification into short-, intermediate-, and long-survivors. Pearson correlation between the predicted and actual OS yielded an R= 0.49. CONCLUSION Multivariate integration of visually quantified tumor characteristics, agnostic to acquisition protocol/equipment, yields an accurate OS surrogate index. Validation of our AI model in the largest centralized glioblastoma imaging dataset, from the ReSPOND consortium, supports its generalizability across diverse patient populations and acquisition settings, potentially contributing to equitable improvements of personalized patient care.

BACKGROUND We evaluated the combination of retifanlimab, INCAGN01876, and FSRT in patients with recurrent GBM. METHODS Phase II, single-center, 2 cohort study (A: single-arm non-surgical cohort; B, two-arm neoadjuvant/surgical cohort). Cohort A patients received pre-FSRT one-time doses of retifanlimab (500mg) and INCAGN01876 (300mg), FSRT (8 Gy x 3 fractions), and post-FSRT 28-day treatment cycles (retifanlimab, day 1; INCAGN01876, days 1, 15). Cohort B patients received pre-surgery doses of retifanlimab + INCAGN01876; subsequently, patients either went directly to resection (Sub-Arm 1) or to FSRT followed by resection (Sub-Arm 2). All patients resumed immunotherapy post-operatively. Primary endpoint: ORR in Cohort A. Data cut-off for this analysis was June 30, 2022. RESULTS Thirty-two evaluable patients: Cohort A, n=16; Cohort B, n=16 (Sub-Arm 1, n=8, Sub-Arm 2, n=8). Median follow-up time: Cohort A, 13.6 months; Cohort B, 8.8 months. Forty-four percent women, median age 64 (IQR, 55–65), 56% MGMT unmethylated. Most common grade 3/4 treatment-related AEs included cerebral edema (25%), lymphopenia (16%), and cognitive disturbance (13%). Efficacy in Cohort A: no objective responses observed, best response of stable disease achieved in 9/16 patients (56%), median PFS 3.9 months (95% CI 2.1 – 6.2 months), median OS 9.8 months (95% CI 8.3 months – not reached [NR]). Efficacy in Cohort B: median PFS NR, median OS 15.1 months (95% CI 8.5 months – NR). Median PFS and OS are longer in Cohort B Sub-Arm 2 compared to Cohort B Sub-Arm 1 (PFS, NR vs. 2.2 months, p = 0.009; OS, NR vs. 8.5 months, p=0.026). Results of tissue and blood-based immune correlative analyses will be presented. CONCLUSIONS The combination of retifanlimab, INCAGN01876, and FSRT is generally well-tolerated in patients with recurrent GBM when administered with or without surgical resection. Survival outcomes in the neoadjuvant cohort are encouraging, largely driven by patients that received neoadjuvant FSRT.

Somatic genomic alterations acquired during GBM growth enhance adaptation of tumor cells to their microenvironment and give rise to molecular heterogeneity. Radiogenomics could facilitate exploration of the underlying pathobiology of tumor growth in specific microenvironments and thereby, promote precision medicine for the patients. We derived radiogenomic signatures of key driver genes and evaluated molecular compositions of tumor groups with predisposition to specific brain regions. Pre-operative multiparametric conventional MRI scans of 357 IDH-wildtype GBM patients with available targeted NGS data were jointly segmented and registered into a common template. We constructed spatial distribution atlases for tumors harboring mutations in driver genes and identified four distinct groups of tumor locations with predilection to the left frontal cingulate region (Group1), right temporal (Group2), right parietal (Group3), and occipital pole (Group4). Evaluation of the differences in molecular features of the tumor groups included: (1) exploring similarities of genomic profiles across all four groups by evaluating cosine similarity metric (CSM) between mutational signatures; (2) quantification of molecular heterogeneity based on Mutant Allele Tumor Heterogeneity (MATH) scores; and (3) inference of the evolutionary trajectories. Groups 1 and 4 were the most different, and Groups 2 and 3 were the most similar tumors, molecularly. The mutational signatures between Groups 1 and 4 revealed a CSM of 0.35. Group1 showed significantly lower MATH score (less heterogeneity) compared to Group4 (p< 0.05). Evaluation of evolutionary patterns suggested NF1 mutation as an early event in Group1, without subsequent gain of function or mutation in EGFR. In contrast, in Group4, EGFR mutations were early events triggering PTEN mutations later in the evolutionary trajectory. Radiogenomic signatures revealed distinct molecular underpinnings for the tumors with predilection towards specific brain regions that may suggest existence of different tumor microenvironments in different brain regions that cause intra- and inter-patient heterogeneity in the molecular tumor composition.

BACKGROUND Diffuse astrocytic glioma are common and aggressive malignant primary brain tumors with grim prognosis. Artificial intelligence (AI) has shown promise across predictive, prognostic, and diagnostic neuro-oncology applications, towards improving patient management. However, clinical translation and deployment are hampered by AI models’ requirements for explicit acceleration cards, which are not typically considered in clinical environments. Here, we seek the execution of AI models in such clinical/low-resource environments, facilitated by mathematical optimizations rather than investing in acceleration cards, and focus on the use case of delineating clinically-relevant regions, namely the entire tumor burden (ETB) and tumor core (TC). METHODS We identified the BraTS20 retrospective cohort of 369 glioma cases, each described by 4 structural multi-parametric MRI (mpMRI) scans (T1,T1Gd,T2,T2-FLAIR). The Generally Nuanced Deep Learning Framework (GaNDLFv0.0.14/PyTorchv1.8.2) was used to train the original AI model (ResUNet), by randomly sampling 40 643 patches from each mpMRI scan. We then investigate the contributions of post-training mathematical optimizations. Quantitative performance evaluation of the original and optimized (GaNDLFv0.0.14/OpenVINOTM-INT8-v2022.1.0) models were based on 125 unseen hold-out cases (BraTS20 validation dataset), using the dice similarity coefficient (DSC), while profiling in a consumer-grade workstation, i.e., typical clinical hardware configuration. RESULTS Negligible delineation performance differences were observed between the original and optimized AI models, for both ETB (DSCoriginal/DSCoptimized= 0.877/0.876) and TC (DSCoriginal/DSCoptimized= 0.773/0.772). However, the optimized model yielded substantial improvements in latency (up to 5.4x faster inference) and 53% less memory footprint. CONCLUSIONS Post-training mathematical optimization of AI models yields substantial gains in latency and memory usage, while maintaining their accuracy. Although we focused on delineation, we anticipate these mathematical optimizations to be applicable in other AI models. Post-training optimization is a promising approach for deploying AI models on consumer-grade CPUs, and hence facilitating their translation in low-resource/clinical environments, potentially contributing to improved patient management, treatment decisions, and response assessment. *author Siddhesh Thakur is an equal contributing first author.

Background. Multi-omic data, i.e., clinical measures, radiomic, and genetic data, capture multi-faceted tumor characteristics, contributing to a comprehensive patient risk assessment. Here, we investigate the additive value and independent reproducibility of integrated diagnostics in prediction of overall survival (OS) in newly diagnosed, treatment-naïve, IDH-wildtype GBM patients, by combining conventional and deep learning methods.Methods. Conventional radiomics and deep learning features were extracted from pre-operative multi-parametric MRI of 516 GBM patients. SVM classifiers were trained on the discovery cohort (n=404) to categorize patient groups of high-risk (OSradiomics) was produced. Multivariate Cox-PH models were generated for the replication cohort, first based on clinical measures solely, and then by layering on radiomics and molecular information.Results. Evaluation of the high-risk and low-risk classifiers in the discovery/replication cohorts revealed AUCs of 0.78 (95%CI:0.70–0.85)/0.75 (95%CI:0.64–0.79) and 0.75 (95%CI: 0.65–0.84)/0.63 (95%CI: 0.52–0.71), respectively. Cox-PH modeling showed a concordance index of 0.65 (95%CI:0.6–0.7) for clinical data, 0.70 (95%CI:0.65–0.75) for clinical and radiomics, 0.72 (95%CI:0.68–0.77) for clinical, MGMT methylation, and radiomics, and 0.75 (95%CI:0.72–0.79) for the combination of all omics, i.e., clinical, MGMT methylation, radiomics, and genomics.Conclusions. This study signifies the value of integrated diagnostics for improved prediction of OS in GBM. Our multi-omic survival prediction tool is easily scalable and can be used for more effective clinical trial stratification.

Histopathologic assessment routinely provides rich microscopic information about tissue structure and disease process. However, the sections used are very thin, and essentially capture only 2D representations of a certain tissue sample. Accurate and robust alignment of sequentially cut 2D slices should contribute to more comprehensive assessment accounting for surrounding 3D information. Towards this end, we here propose a two-step diffeomorphic registration approach that aligns differently stained histology slides to each other, starting with an initial affine step followed by estimating a deformation field. It was quantitatively evaluated on ample (n = 481) and diverse data from the automatic non-rigid histological image registration challenge, where it was awarded the second rank. The obtained results demonstrate the ability of the proposed approach to robustly (average robustness = 0.9898) and accurately (average relative target registration error = 0.2%) align differently stained histology slices of various anatomical sites while maintaining reasonable computational efficiency (&lt, 1 min per registration). The method was developed by adapting a general-purpose registration algorithm designed for 3D radiographic scans and achieved consistently accurate results for aligning high-resolution 2D histologic images. Accurate alignment of histologic images can contribute to a better understanding of the spatial arrangement and growth patterns of cells, vessels, matrix, nerves, and immune cell interactions.

Background Pituicytoma is a rare tumor of the sella and suprasellar region. It is common for these tumors to be misdiagnosed radiographically as pituitary adenomas, meningiomas, and craniopharyngiomas. Histologically, pituicytomas are also known to show variable morphology. These factors often complicate intraoperative consultation, especially when tissue is limited. Methods A database search (January 1990-June 2019) identified 13 surgically resected pituicytomas that were sent for intraoperative consultation. The intraoperative cytology was reviewed by two pathologists, and both squash preparations/smears and touch preparations were included. Results The cytological features of pituicytomas were variable. The cytoplasm ranged from fibrillary to fine and wispy. The nuclei were round to ovoid with occasional tumors showing spindled morphology. Small nucleoli were seen in all tumors, and chromatin was fine to vesicular. While squash preparations were cellular and revealed cohesive groups of tumor cells, touch preparations were often paucicellular with the exception of one tumor. The cytological features of pituicytomas are compared to those of other tumors with similar radiographic appearances. Conclusions Our findings illustrate the range of cytological features of these rare tumors and highlight the difficulty in making a definitive intraoperative diagnosis. Pituicytomas should be considered in the differential of a sellar/suprasellar lesion with glial and meningioma-like cytological features.

OBJECTIVEIntraoperative molecular imaging with tumor-targeted fluorescent dyes can enhance resection rates. In contrast to visible-light fluorophores (e.g., 5-aminolevulinic-acid), near-infrared (NIR) fluorophores have increased photon tissue penetration and less contamination from tissue autofluorescence. The second-window ICG (SWIG) technique relies on passive accumulation of indocyanine green (ICG) in neoplastic tissues. OTL38, conversely, targets folate receptor overexpression in nonfunctioning pituitary adenomas. In this study, we compare the properties of these 2 modalities for NIR imaging of pituitary adenomas to better understand the potential for NIR imaging in neurosurgery.METHODSA total of 39 patients with pituitary adenomas were enrolled between June 2015 and January 2018 in 2, sequential, IRB-approved studies. Sixteen patients received systemic ICG infusions 24 hours prior to surgery, and another 23 patients received OTL38 infusions 2–3 hours prior to surgery. NIR fluorescence signal-to-background ratio (SBR) was recorded during and after resection. Immunohistochemistry was performed on the 23 adenomas resected from patients who received OTL38 to assess expression of folate receptor–alpha (FRα).RESULTSAll 16 adenomas operated on after ICG administration demonstrated strong NIR fluorescence (mean SBR 4.1 ± 0.69 [SD]). There was no statistically significant difference between the 9 functioning and 7 nonfunctioning adenomas (p = 0.9). After administration of OTL38, the mean SBR was 1.7 ± 0.47 for functioning adenomas, 2.6 ± 0.91 for all nonfunctioning adenomas, and 3.2 ± 0.53 for the subset of FRα-overexpressing adenomas. Tissue identification with white light alone for all adenomas demonstrated 88% sensitivity and 90% specificity. SWIG demonstrated 100% sensitivity but only 29% specificity for both functioning and nonfunctioning adenomas. OTL38 was 75% sensitive and 100% specific for all nonfunctioning adenomas, but when assessment was limited to the 9 FRα-overexpressing adenomas, the sensitivity and specificity of OTL38 were both 100%.CONCLUSIONSIntraoperative imaging with NIR fluorophores demonstrates highly sensitive detection of pituitary adenomas. OTL38, a folate-receptor–targeted fluorophore, is highly specific for nonfunctioning adenomas but has no utility in functioning adenomas. SWIG, which relies on passive diffusion into neoplastic tissue, is applicable to both functioning and nonfunctioning pituitary adenomas, but it is less specific than targeted fluorophores. Thus, targeted and nontargeted NIR fluorophores play important, yet distinct, roles in intraoperative imaging. Selectively and intelligently using either agent has the potential to greatly improve resection rates and outcomes for patients with intracranial tumors.

Malignant peripheral nerve sheath tumors contain loss of histone H3K27 trimethylation (H3K27me3) due to driver mutations affecting the polycomb repressive complex 2 (PRC2). Consequently, loss of H3K27me3 staining has served as a diagnostic marker for this tumor type. However, recent reports demonstrate H3K27me3 loss in numerous other tumors, including some in the differential diagnosis of malignant peripheral nerve sheath tumor. Since these tumors lose H3K27me3 through mechanisms distinct from PRC2 loss, we set out to determine whether loss of dimethylation of H3K27, which is also catalyzed by PRC2, might be a more specific marker of PRC2 loss and malignant peripheral nerve sheath tumor. Using mass spectrometry, we identify a near complete loss of H3K27me2 in malignant peripheral nerve sheath tumors and cell lines. Immunohistochemical analysis of 72 malignant peripheral nerve sheath tumors, seven K27M-mutant gliomas, 43 ependymomas, and 10 Merkel cell carcinomas demonstrates that while H3K27me3 loss is common across these tumor types, H3K27me2 loss is limited to malignant peripheral nerve sheath tumors and is highly concordant with H3K27me3 loss (33/34 cases). Thus, increased specificity does not come at the cost of greatly reduced sensitivity. To further compare H3K27me2 and H3K27me3 immunohistochemistry, we investigated 42 melanomas and 54 synovial sarcomas, histologic mimics of malignant peripheral nerve sheath tumor with varying degrees of H3K27me3 loss in prior reports. While global H3K27me3 loss was not seen in these tumors, weak and limited H3K27me3 staining was common. By contrast, H3K27me2 staining was more clearly retained in all cases, making it a superior binary classifier. This was confirmed by digital image analysis of stained slides. Our findings indicate that H3K27me2 loss is highly specific for PRC2 loss and that PRC2 loss is a rarer phenomenon than H3K27me3 loss. Consequently, H3K27me2 loss is a superior diagnostic marker for malignant peripheral nerve sheath tumor.

Young adults with isocitrate-dehydrogenase wild-type (IDH-WT) glioblastoma (GBM) represent a rare, understudied population compared to pediatric high-grade glioma, IDH-mutant GBM, or IDH-WT GBM in older patients. We aimed to explore the prognostic impact of epidermal growth factor receptor copy number gain (EGFR CN gain), one of the most common genetic alterations in IDH-WT glioma, in young adults with IDH-WT GBM. We performed a retrospective cohort study of patients 18–45 years old with newly diagnosed, IDH-WT GBM whose tumors underwent next-generation sequencing at our institution between 2014 and 2018. The impact of EGFR CN gain on time to tumor progression (TTP) and overall survival (OS) was assessed. A validation cohort of patients 18–45 years old with IDH-WT GBM was analyzed from The Cancer Genome Atlas (TCGA). Ten of 28 patients (36%) from our institution had EGFR CN gain, which was associated with shorter TTP (median 6.5 vs. 11.9 months; p = 0.06) and OS (median 16.3 vs. 23.5 months; p = 0.047). The negative prognostic impact of EGFR CN gain on OS persisted in a multivariate model (HR 6.40, 95% CI 1.3–31.0, p = 0.02). In the TCGA cohort (N = 43), EGFR CN gain was associated with shorter TTP and worse OS, although these did not reach statistical significance (TTP, median 11.5 vs. 14.4 months, p = 0.18; OS, median 23.6 vs. 27.8 months; p = 0.18). EGFR CN gain may be associated with inferior outcomes in young adults with newly diagnosed, IDH-WT GBM, suggesting a potential role for targeting EGFR in this population.

Background Stereotactic needle biopsy provides a minimally invasive option for the diagnosis of intracranial lesions but is limited by inconclusive diagnoses on frozen pathology. For rapid pathology, 5-aminovelunic acid and sodium fluorescein have previously demonstrated potential as diagnostic adjuvants. Stereotactic biopsy with near-infrared (NIR) fluorophores has not been reported. We identified 5 representative cases using NIR fluorescent dye indocyanine green (ICG) administered in a high dose, delayed manner. Methods Five patients underwent second window indocyanine green (SWIG)-guided stereotactic biopsy for diagnosis of suspected glioma or tumor recurrence. Up to 5 mg/kg ICG was administered approximately 24 hours prior to surgery. Biopsies were conducted in the standard fashion, targeting regions of suspected tumor using intraoperative frameless navigation. Samples were examined intraoperatively under standard visible light and for fluorescence using conventional NIR imaging platforms. Findings were correlated with frozen and final tumor pathology for all cases. Results A total of 10 biopsy specimens were obtained. Three did not fluoresce and did not demonstrate tumor on preliminary or final pathology, including a non-gadolinium-enhancing sample taken proximal to the final target. The remaining 7 fluoresced, of which 6 contained tumor and 1 contained necrosis. Fluorescence was also noted in a patient with radiation treatment effect. Overall fluorescence characteristics were highly concordant with preliminary and final diagnoses. Conclusions SWIG provides rapid intraoperative confirmation of pathologic brain tissue by permeating neoplastic or inflammatory brain tissue via a mechanism similar to that of gadolinium enhancement. SWIG-guided stereotactic biopsy can improve surgical efficiency by enhancing confidence in acquisition of target tissue.

Elevated levels of choline are generally emphasized as marker of increased cellularity and cell membrane turnover in gliomas. In this study, we investigated the incidence rate of lack of choline/creatine and choline/water elevation in a population of grade I–III gliomas. A cohort of 41 patients with histopathologically confirmed gliomas underwent multi-voxel proton magnetic resonance spectroscopy on a 3 T magnetic resonance system prior to treatment. Peak areas for choline and myoinositol were measured from all voxels that exhibited hyperintensity on fluid-attenuated inversion recovery images and were normalized to creatine and unsuppressed water from each voxel. The average metabolite/creatine and metabolite/water ratios from these voxels were then computed. Similarly, average metabolite ratios were computed from normal brain parenchyma. Gliomas were considered for lack of choline elevation when choline/creatine and choline/water ratios from neoplastic regions were less than those from normal brain parenchyma regions. Six of 41 (14.6%) grade I–III gliomas showed lack of elevation for choline/creatine and choline/water ratios compared to normal brain parenchyma. Four of these six gliomas also demonstrated elevated levels of myoinositol/creatine ratio. All other gliomas ( n = 35) had elevated choline levels from neoplastic regions relative to normal parenchyma. The sensitivity of choline/creatine or choline/water in determining a grade I–III glioma was 85.4%. These findings suggest that a lack of choline/creatine or choline/water elevation may be seen in some gliomas and low choline levels should not prevent us from considering the possibility of a grade I–III glioma.

Bispecific T-cell engagers (BiTEs) are bispecific antibodies that redirect T cells to target antigen-expressing tumors. BiTEs can be secreted by T cells through genetic engineering and perform anti-tumor activity. We hypothesized that BiTE-secreting T cells could be a valuable T cell-directed therapy in solid tumors, with distinct properties in mono- or multi-valent strategies incorporating chimeric antigen receptor (CAR) T cells. Glioblastomas represent a good model for solid tumor heterogeneity and represent a significant therapeutic challenge. We detected expression of tumor-associated epidermal growth factor receptor (EGFR), EGFR variant III (EGFRvIII), and interleukin-13 receptor alpha 2 (IL13Rα2) on glioma tissues and glioma cancer stem cells. These antigens formed the basis of a multivalent approach, using a conformation-specific tumor-related EGFR targeting antibody (806) and Hu08, an IL13Rα2-targeting antibody, as the scFvs to generate new BiTE molecules. Compared with 806CAR T cells and Hu08CAR T cells, BiTE T cells demonstrated prominent activation, cytokine production, and cytotoxicity in response to target-positive gliomas. Superior response activity was also demonstrated in BiTE secreting bivalent targeting T cells compared with bivalent targeting CAR T cells, which significantly delayed tumor growth in a glioma mouse model. In summary, BiTEs secreted by mono- or multi- valent targeting T cells have potent anti-tumor activity in vitro and in vivo with significant sensitivity and specificity, demonstrating a promising strategy in solid tumor therapy.

PURPOSE Decision making about the best course of treatment for glioblastoma patients becomes challenging when a new enhancing lesion appears in the vicinity of the surgical bed on follow-up MRI (after maximal safe tumor resection and chemoradiation), raising concerns for tumor progression (TP). Literature indicates 30-50% of these new lesions describe primarily treatment-related changes (TRC). We hypothesize that quantitative analysis of specific and sensitive features extracted from multi-parametric MRI (mpMRI) via machine learning (ML) techniques may yield non-invasive imaging signatures that distinguish TP from TRC and facilitate better treatment personalization. METHODS We have generated an ML model on a retrospective cohort of 58 subjects, and prospectively evaluated on an independent cohort of 58 previously unseen patients who underwent second resection for suspicious recurrence and had availability of advanced mpMRI (T1, T1-Gd, T2, T2-FLAIR, DTI, DSC). The features selected by our retrospective model, representing principal components analysis of intensity distributions, morphological, statistical, and texture descriptors, were extracted from the mpMRI of the prospective cohort. Integration of these features revealed signatures distinguishing between TP, mixed response, and TRC. Independently, a board-certified neuropathologist evaluated the resected tissue by blindly classifying it in the above three categories, based on mitotic figures, pseudopalisading necrosis, geographic necrosis, dystrophic calcification, vascular changes, and Ki67. RESULTS Tissues classified as TRC by the neuropathologist were associated with imaging phenotypes of lower angiogenesis (DSC-derived features), lower cellularity (DTI-derived features), and higher water concentration (T2, T2-FLAIR features). Our ML model characterized TP with 78% accuracy (sensitivity:86%, specificity:70%, AUC:0.80 (95%CI, 0.68-0.92)) and TRC with 81% accuracy (sensitivity:80%, specificity:81%, AUC:0.87 (95%CI, 0.72-1.00)). CONCLUSION Our proposed ML model reveals distinct non-invasive markers of TP and TRC, directly associated with histopathological changes in prospective glioblastoma patients. Reliable stratification of TP and TRC entities may help to noninvasively determine whether the course of treatment should change.

PURPOSE Multi-parametric MRI and artificial intelligence (AI) methods were previously used to predict peritumoral neoplastic cell infiltration and risk of future recurrence in glioblastoma, in single-institution studies. We hypothesize that important characteristics of peritumoral tissue heterogeneity captured, engineered/selected, and quantified by these methods relate to predictions generalizable in the multi-institutional ReSPOND (Radiomics Signatures for PrecisiON Diagnostics) consortium. METHODS To support further development, generalization, and clinical translation of our proposed method, we trained the AI model on a retrospective cohort of 29 de novo glioblastoma patients from the Hospital of the University of Pennsylvania (UPenn) (Male/Female:20/9, age:22-78 years) followed by evaluation on a prospective multi-institutional cohort of 84 glioblastoma patients (Male/Female:51/33, age:34-89 years) from Case Western Reserve University/University Hospitals (CWRU/UH, 25), New York University (NYU, 13), Ohio State University (OSU, 13), University Hospital Río Hortega (RH, 2), and UPenn (31). Features extracted from pre-resection MRI (T1, T1-Gd, T2, T2-FLAIR, ADC) were used to build our model predicting the spatial pattern of subsequent tumor recurrence. These predictions were evaluated against regions of pathology-confirmed post-resection recurrence. RESULTS Our model predicted the locations that later harbored tumor recurrence with sensitivity 83%, AUC 0.83 (99% CI, 0.73-0.93), and odds ratio 7.23 (99% CI, 7.09-7.37) in the prospective cohort. Odds ratio (99% CI)/AUC(99% CI) per institute were: CWRU/UH, 7.8(7.6-8.1)/0.82(0.75-0.89); NYU, 3.5(3.3-3.6)/0.84(074-0.93); OSU, 7.9(7.6-8.3)/0.8(0.67-0.94); RH, 22.7(20-25.1)/0.94(0.27-1); UPenn, 7.1(6.8-7.3)/0.83(0.76-0.91). CONCLUSION This is the first study that provides relatively extensive multi-institutional validated evidence that AI can provide good predictions of peritumoral neoplastic cell infiltration and future recurrence, by dissecting the MRI signal heterogeneity in peritumoral tissue. Our analyses leveraged the unique dataset of the ReSPOND consortium, which aims to develop and evaluate AI-based biomarkers for individualized prediction and prognostication, by moving from single-institution studies to generalizable, well-validated multi-institutional predictive biomarkers.

Accurate glioma classification is essential for optimal patient care, and requires integration of histological and immunohistochemical findings with molecular features, in the context of imaging and demographic information. This paper will introduce classic histologic features of gliomas in contrast to nonneoplastic brain parenchyma, describe the basic clinical algorithm used to classify infiltrating gliomas, and demonstrate how the classification is reflected in the diagnostic reporting structure. Key molecular features include IDH mutational status and 1p/19q codeletion. In addition, molecular changes may indicate poor prognosis despite lower grade histology, such as findings consistent with two grade IV infiltrating gliomas: molecular glioblastoma and diffuse midline gliomas. Detailed molecular characterization aids in optimization of treatment with the goal of improved patient outcomes.

This article has been withdrawn due to an error that caused the article to be duplicated. The definitive version of this article is published under DOI 10.1093/neuros/nyab288.

AI-based methods have shown great promise in a variety of biomedical research fields, including neurooncologic imaging. For example, machine learning methods have offered informative predictions of overall survival (OS) and progression-free survival (PFS), differentiation between pseudoprogression (PsP) and progressive disease (PD), and estimation of mutational status from imaging data. Despite their promise, AI, and especially the emerging deep learning (DL) methods, are challenged by several factors, including imaging heterogeneity across scanners and lack of sufficiently large and diverse training datasets, which limits their reproducibility and general acceptance. These challenges prompted the development of the ReSPOND (Radiomics Signatures for PrecisiON Diagnostics) consortium on glioblastoma, a growing effort to bring together a community of researchers sharing imaging, demographic, clinical and (currently) limited molecular data in order to address the following aims: 1) pool and harmonize data across diverse hospitals and patient populations worldwide; 2) derive robust and generalizable AI models for prediction of (initially) OS, PFS, PsP vs. PD, and recurrence; 3) test these predictive models across multiple sites. In its first phase, ReSPOND aims to pool together approximately 3,000MRI scans (from 10institutions plus TCIA), along with demographics, KPS, and (for a subset) MGMT/IDH1 status. We present initial results testing the generalization of a previously trained model of OS on 505Penn datasets to 2independent cohorts from Case Western Reserve University and University Hospitals (N=44), and Penn (N=67). The results indicate good generalization, with correlation coefficients between OS/predicted-OS between 0.25 to 0.5, depending on variable availability, which is comparable to cross-validated accuracy previously obtained from the training set itself (N=505). Additional preliminary studies evaluating prediction of future recurrence from baseline pre-operative scans in de novo patients (Penn model applied to CWR) indicated potential for guiding targeted dose escalation and supra-total resection (excellent predictions in 6/12 patients, modest in 1/12, and poor in 5/12).

Pediatric ependymoma is a devastating brain cancer marked by its relapsing pattern and lack of effective chemotherapies. This shortage of treatments is partially due to limited knowledge about ependymoma tumorigenic mechanisms. Although there is evidence that ependymoma originates in radial glia, the specific pathways underlying the progression and metastasis of these tumors are unknown. By means of single-cell transcriptomics, immunofluorescence, and in situ hybridization, we show that the expression profile of tumor cells from pediatric ependymomas in the posterior fossa is consistent with an origin in LGR5+ stem cells. Tumor stem cells recapitulate the developmental lineages of radial glia in neurogenic niches, promote an inflammatory microenvironment in cooperation with microglia, and upon metastatic progression initiate a mesenchymal program driven by reactive gliosis and hypoxia-related genes. Our results uncover the cell ecosystem of pediatric posterior fossa ependymoma and identify WNT/β-catenin and TGF-β signaling as major drivers of tumorigenesis for this cancer.

Glioblastoma (‘GBM’) is the most aggressive type of primary malignant adult brain tumor, with very heterogeneous radiographic, histologic, and molecular profiles. A growing body of advanced computational analyses are conducted towards further understanding the biology and variation in glioblastoma. To address the intrinsic heterogeneity among different computational studies, reference standards have been established to facilitate both radiographic and molecular analyses, e.g., anatomical atlas for image registration and housekeeping genes, respectively. However, there is an apparent lack of reference standards in the domain of digital pathology, where each independent study uses an arbitrarily chosen slide from their evaluation dataset for normalization purposes. In this study, we introduce a novel stain normalization approach based on a composite reference slide comprised of information from a large population of anatomically annotated hematoxylin and eosin (‘H&E’) whole-slide images from the Ivy Glioblastoma Atlas Project (‘IvyGAP’). Two board-certified neuropathologists manually reviewed and selected annotations in 509 slides, according to the World Health Organization definitions. We computed summary statistics from each of these approved annotations and weighted them based on their percent contribution to overall slide (‘PCOS’), to form a global histogram and stain vectors. Quantitative evaluation of pre- and post-normalization stain density statistics for each annotated region with PCOS\(\,>\,0.05\)% yielded a significant (largest p\(\,=\,\)0.001, two-sided Wilcoxon rank sum test) reduction of its intensity variation for both ‘H’ & ‘E’. Subject to further large-scale evaluation, our findings support the proposed approach as a potentially robust population-based reference for stain normalization.

Glioblastoma (GBM) is the deadliest form of brain cancer and recurs uniformly. Despite aggressive treatment with maximal safe surgical resection, adjuvant radiation with temozolomide chemotherapy, and alternating electrical field therapy, median survival for newly diagnosed GBM remains

Metastatic medullary thyroid cancer (MTC) is a rare but often aggressive thyroid malignancy with a 5-year survival rate of less than 40% and few effective therapeutic options. Adoptive T cell immunotherapy using chimeric antigen receptor (CAR)-modified T cells (CAR Ts) is showing encouraging results in the treatment of cancer, but development is challenged by the availability of suitable target antigens. We identified glial-derived neurotrophic factor (GDNF) family receptor alpha 4 (GFRα4) as a putative antigen target for CAR-based therapy of MTC. We show that GFRα4 is highly expressed in MTC, in parafollicular cells within the thyroid from which MTC originates, and in normal thymus. We isolated two single-chain variable fragments (scFvs) targeting GFRα4 isoforms a and b by antibody phage display. CARs bearing the CD3ζ and the CD137 costimulatory domains were constructed using these GFRα4-specific scFvs. GFRα4-specific CAR Ts trigger antigen-dependent cytotoxicity and cytokine production in vitro, and they are able to eliminate tumors derived from the MTC TT cell line in an immunodeficient mouse xenograft model of MTC. These data demonstrate the feasibility of targeting GFRα4 by CAR T and support this antigen as a promising target for adoptive T cell immunotherapy and other antibody-based therapies for MTC., Graphical Abstract, Metastatic medullary thyroid carcinoma (MTC) is a rare but aggressive malignancy with no curative options. By sequencing mRNA from a patient’s tumor, Siegel and colleagues identified and characterized GFRα4 as a promising tumor target antigen and demonstrated that CAR T immunotherapy targeting this antigen controlled MTC in tumor xenograft models.