Yoon, Hyunsoo

Abstract Introduction Distinguishing between tumor and treatment effect in post-treatment glioma, although crucial for clinical management, is difficult because contrast-enhancing regions are mixtures of recurrent tumor and reactive tissue, and definitive histopathological criteria do not exist. This study disentangles the marked intra-tumoral heterogeneity in the treatment-recurrent setting by developing an unsupervised framework to algorithmically categorize intraoperative MRI-localized biopsies into three clinically-relevant tissue clusters based on joint analysis of RNA sequencing and histopathological data. Methods A retrospective cohort of 84 MRI-localized biopsies from 37 patients with post-treatment recurrent glioblastoma underwent mRNA extraction and quantification via PLATEseq protocol. For 48 of 84 biopsies, a neighboring piece of tissue underwent quantitative histopathology based on labeling index (LI) for SOX2, CD68, NeuN, Ki67, and H&E. Correlation between LIs and gene expression for these 48 samples was performed. Genes significantly correlated (p<0.05) with ≥1 marker were used for hierarchical clustering of correlation matrix, identifying three mutually-exclusive tissue-specific gene sets. These sets were then used to perform ssGSEA to categorize each of 84 biopsies into one of three tissue types. Results Correlation analysis identified 7779 genes significantly correlated with ≥1 histopathological marker. Clustering revealed three gene sets associated with specific markers: SetA-3688 genes associated with SOX2/Ki67/H&E; SetB-2418 genes associated with CD68; SetC-1673 genes associated with NeuN. ssGSEA using these sets categorized each biopsy into one of three tissue types: 27 biopsies enriched in SetA, 28 in SetB, and 29 in SetC. Conclusions Using MRI-localized biopsies with both RNAseq and histopathological data, this algorithmic approach allows development of three orthogonal tissue-specific gene sets that can be applied to characterize the heterogeneity in post-treatment recurrent glioma: SetA: genes correlated with SOX2/Ki67/H&E, representing recurrent tumor; SetB: genes correlated with CD68 (microglia) representing reactive tissue consistent with treatment effect; SetC: genes correlated with NeuN (neurons), representing infiltrated brain.

Abstract BACKGROUND Conventional magnetic resonance imaging (MR) guides patient care in GBM. However, there is mounting awareness that MR enhancement is non-specific reflecting either tumor progression or non-tumoral inflammatory changes. Histological evaluation of GBM is held as the gold standard for disease assessment. However, the invasiveness of this methodology and the sample sparsity limit its usefulness. Methods to infer histological underpinnings of MRI are needed to improve clinical care. METHODS A transfer learning mixed effects model based on T1Gd and FLAIR MR voxel based image features trained and cross-validated to predict FPKM values of CD68, CASP3, CD8A and IL13RA2. Training data included RNAseq from 38 image-localized biopsies from 15 newly-diagnosed GBM patients. These models were then applied to two independent patients, chosen based on therapy and the quality of image registration, at three different time points, just prior to receiving IL13RA2 targeted CAR-T therapy for rGBM, after 2 cycles and after 4 cycles, resulting in a voxel-based prediction of the relative expression levels of these genes. RESULTS Cross-validation of the proposed machine learning models demonstrated a Pearson Correlation coefficient between predicted and observed FPKM values of 0.89 (CD68), 0.92 (CASP3), 0.93 (CD8A), and 0.88 (IL13RA2). When the models were applied to the two CAR-T patients, CD68-modeled expression levels were seen to increase throughout therapy for one patient, while remaining stable for the other. Survival from first CAR-T infusion was, respectively, 412 and 83 days suggesting that the increased CD68 was indicative of therapeutic effectivity. Further, the spatial activity predictions were consistent with expected therapeutic action as the correlation coefficient between CASP3 and the product of IL13R2 and CD8A expression was 0.81, suggesting the T-cells targeting IL13RA2 were inducing cell death. CONCLUSIONS Preliminary results with our model highlights the potential of spatially resolved radiomic maps to provide insight into regional therapeutic effectivity.

Abstract Recurrent glioblastoma is challenging to distinguish from so called “treatment effect” on routine clinical imaging. Further, within tumor heterogeneity reveals that some regions can be histologically dominated by tumor progression whilst others can be dominated by secondary effects of treatment response. Apparent tumor progression on MRI can be very difficult to manage clinically as it is unclear the degree to which the imaging changes are actually tumor progression vs response to treatment (including inflammatory response and necrosis). In this analysis, we study a unique cohort of patients for whom image localized-biopsies reveal heterogeneity in response vs progression. Our dataset included 70 biopsy samples from 32 patients with GBM each histolopatholgically characterized for tumor abundance vs immune infiltrate. Six multiparametric MRI contrasts were available, including T1, T1gd, T2, FLAIR, SWI, and ADC. Images were co-registered. Radiomic (statistical + texture) features were extracted from the region of six image contrasts locally matched with each biopsy sample. Machine learning models were built to predict each biomarker using radiomic features. Leave-one-out cross validation was used to evaluate the prediction accuracy. Radiomic features were found to be informative to the prediction of biomarkers. ANOVA tests show significant improvement of using radiomic features compared with the null model. The prediction accuracy was higher when considering the biomarkers on a binary scale using the median as the cutoff than on a numerical scale. Spatially-informed radiomics models for tumor progression vs treatment effect are possible and can play an instrumental role in navigating confounding imaging changed common during treatment progression.

Abstract BACKGROUND An important challenge in radiomics research is reproducibility. Images are collected on different image scanners and protocols, which introduces significant variability even for the same type of image across institutions. In the present proof-of-concept study, we address the reproducibility issue by using domain adaptation – an algorithm that transforms the radiomic features of each new patient to align with the distribution of features formed by the patient samples in a training set. METHOD Our dataset included 18 patients in training with a total of 82 biopsy sample. The pathological tumor cell density was available for each sample. Radiomic (statistical + texture) features were extracted from the region of six image contrasts locally matched with each biopsy sample. A Gaussian Process (GP) classifier was built to predict tumor cell density using radiomic features. Another 6 patients were used to test the training model. These patients had a total of 31 biopsy samples. The images of each test patient were purposely normalized using a different approach, i.e., using the CSF instead of the whole brain as the reference. This was to mimic the practical scenario of image source discrepancy between different patients. Domain adaptation was applied to each test patient. RESULTS Among the 18 training patients, the leave-one-patient-out cross validation accuracy is 0.81 AUC, 0.78 sensitivity, and 0.83 specificity. When the trained model was applied to the 6 test patients (purposely normalized using a different approach than that of the training data), the accuracy dramatically reduced to 0.39 AUC, 0.08 sensitivity, and 0.61 specificity. After using domain adaption, the accuracy improved to 0.68 AUC, 0.62 sensitivity, and 0.72 specificity. CONCLUSION We provide candidate enabling tools to address reproducibility in radiomics models by using domain adaption algorithms to account for discrepancy of the images between different patients.

Abstract INTRODUCTION The quantification of intratumoral heterogeneity – through radiomics-based approaches - can help resolve the regionally distinct genetic drug targets that may co-exist within a single Glioblastoma (GBM) tumor. While this offers potential diagnostic value under the paradigm of individualized oncology, clinical decision-making must also consider the degree of uncertainty associated with each model. In this study, we evaluate the performance of a novel machine-learning (ML) algorithm, called Gaussian Process (GP) modeling, that can quantify the impact of multiple sources of uncertainty in ML model development and prediction accuracy, including variabilities in the copy number measurement, radiomics features, training sample characteristics, and training sample size. METHOD We collected 95 image-localized biopsies from 25 primary GBM patients. We coregistered stereotactic locations with preoperative multi-parametric MRI features (conventional MRI, DSC perfusion, Diffusion Tensor Imaging) to generate spatially matched pairs of MRI and copy number variants (CNV) for for each biopsy. We developed a Gaussian Process (GP) model to predict CNV for Epidermal Growth Factor Receptor (EGFR) based on MRI radiomic features in each patient. We used leave-one-patient-out cross validation to quantify prediction accuracy and model uncertainty. Spatial prediction and uncertainty (p-value) maps were overlaid to help visualize regional genetic variation of EGFR and uncertainty of the radiomic predictions. RESULT: The initial GP radiomics model for EGFR amplification (CNV > 3.5) produced a sensitivity of 0.8 and specificity of 0.8. Samples/regions associated with high uncertainty (p-value >0.05) correlated with either 1) extrapolation of radiomic features from the training set-defined feature space or 2) insufficient training samples in the feature space. CONCLUSION We present a ML-based model that quantifies spatial genetic heterogeneity in GBM, while also estimating model uncertainties that result from multi-source data variabilities. This approach lays the groundwork for prospective clinical integration of modeling-based diagnostic approaches in the paradigm of individualized medicine.

Abstract High-grade gliomas (HGGs) nearly always recur after standard initial treatment, and the resulting mixture of recurrent tumor and treatment-induced reactive changes presents major diagnostic challenges. Anatomical imaging, such as MRI, cannot adequately distinguish progressive disease from treatment effect (pseudo-progression). Furthermore, there is marked intra-tumoral heterogeneity, such that some areas of a tumor may demonstrate necrotic treatment effect and others frank recurrence. Due to this difficulty reliably differentiating between these two clinical findings, analytic methods using multiple modalities are necessary to further our understanding of this disease process. To this end, we sought to correlate radiographic, histopathologic and molecular features of surgically sampled post-treatment suspected recurrence to identify markers distinguishing tumor growth from treatment effect. We performed Stimulated Raman Histology (SRH) imaging and highly multiplexed RNA-sequencing (PLATE-seq) on 84 MRI-localized biopsies from 39 patients with clinically suspected recurrent HGG. The SRH images were classified as recurrent tumor or gliotic/reactive tissue using a convolutional neural network trained on an independent cohort including a large set of recurrent HGG, and an automated cell-counting algorithm was used to quantify cellularity from the SRH image of each sample. Differential gene expression analysis of the PLATE-seq data was used to identify gene sets that distinguish recurrent tumor from treatment effect, and single sample gene set variation analysis (GSVA) was used to further assess the molecular and cellular composition of each MRI-localized sample. The histopathologic and molecular features of each sample were also correlated with the MRI features of the corresponding biopsy sites, and this data is currently being used to train machine learning models that predict the distribution of recurrent tumor and treatment-induced reactive changes within a patient’s radiographic lesion. These predictive radiomic models will help to guide neurosurgical sampling, and improve our ability to monitor glioma progression and response to therapy.