Aikaterini Kotrotsou

Radiomics in Brain Tumors: An Emerging Technique for Characterization of Tumor Environment

The role of radiomics in the diagnosis, monitoring, and therapy planning of brain tumors is becoming increasingly clear. Incorporation of quantitative approaches in radiology, in combination with increased computer power, offers unique insights into macroscopic tumor characteristics and their direct association with the underlying pathophysiology. This article presents the most recent findings in radiomics and radiogenomics with respect to identifying potential imaging biomarkers with prognostic value that can lead to individualized therapy. In addition, a brief introduction to the concept of big data and its significance in medicine is presented.

139 Clinically Applicable and Biologically Validated MRI Radiomic Test Method Predicts Glioblastoma Genomic Landscape and Survival.

INTRODUCTION Imaging is the modality of choice for noninvasive characterization of biological tissue and organ systems; imaging serves as early diagnostic tool for most disease processes and is rapidly evolving, thus transforming the way we diagnose and follow patients over time. A vast number of cancer imaging characteristics have been correlated to underlying genomics; however, none have established causality. Therefore, our objectives were to test if there is a causal relationship between imaging and genomic information; and to develop a clinically relevant radiomic pipeline for glioblastoma molecular characterization. METHODS Functional validation was performed using a prototypic in vivo RNA-interference-based orthotopic xenograft mouse model. The automated pipeline collects 4800 MRI-derived texture features per tumor. Using univariate feature selection and boosted tree predictive modeling, a patient-specific genomic probability map was derived and patient survival predicted (The Cancer Genome Atlas/MD Anderson data sets). RESULTS Data demonstrated a significant xenograft to human association (area under the curve [AUC] 84%, P < .001). Further, epidermal growth factor receptor amplification (AUC 86%, P < .0001), O-methylguanine-DNA-methyltransferase methylation/expression (AUC 92%, P = .001), glioblastoma molecular subgroups (AUC 88%, P = .001), and survival in 2 independent data sets (AUC 90%, P < .001) was predicted. CONCLUSION Our results for the first time illustrate a causal relationship between imaging features and genomic tumor composition. We present a directly clinically applicable analytical imaging method termed Radiome Sequencing to allow for automated image analysis, prediction of key genomic events, and survival. This method is scalable and applicable to any type of medical imaging. Further, it allows for human-mouse matched coclinical trials, in-depth end point analysis, and upfront noninvasive high-resolution radiomics-based diagnostic, prognostic, and predictive biomarker development.

Radiomic Texture Analysis Mapping Predicts Areas of True Functional MRI Activity

Individual analysis of functional Magnetic Resonance Imaging (fMRI) scans requires user-adjustment of the statistical threshold in order to maximize true functional activity and eliminate false positives. In this study, we propose a novel technique that uses radiomic texture analysis (TA) features associated with heterogeneity to predict areas of true functional activity. Scans of 15 right-handed healthy volunteers were analyzed using SPM8. The resulting functional maps were thresholded to optimize visualization of language areas, resulting in 116 regions of interests (ROIs). A board-certified neuroradiologist classified different ROIs into Expected (E) and Non-Expected (NE) based on their anatomical locations. TA was performed using the mean Echo-Planner Imaging (EPI) volume, and 20 rotation-invariant texture features were obtained for each ROI. Using forward stepwise logistic regression, we built a predictive model that discriminated between E and NE areas of functional activity, with a cross-validation AUC and success rate of 79.84% and 80.19% respectively (specificity/sensitivity of 78.34%/82.61%). This study found that radiomic TA of fMRI scans may allow for determination of areas of true functional activity, and thus eliminate clinician bias.

Radiomic Phenotyping in Brain Cancer to Unravel Hidden Information in Medical Images

Abstract Radiomics is a new area of research in the field of imaging with tremendous potential to unravel the hidden information in digital images. The scope of radiology has grown exponentially over the last two decades; since the advent of radiomics, many quantitative imaging features can now be extracted from medical images through high-throughput computing, and these can be converted into mineable data that can help in linking imaging phenotypes with clinical data, genomics, proteomics, and other “omics” information. In cancer, radiomic imaging analysis aims at extracting imaging features embedded in the imaging data, which can act as a guide in the disease or cancer diagnosis, staging and planning interventions for treating patients, monitor patients on therapy, predict treatment response, and determine patient outcomes.

A co-clinical radiogenomic validation study - Conserved magnetic resonance radiomic appearance of Periostin expressing Glioblastoma in patients and xenograft models

Purpose: Radiomics is the extraction of multidimensional imaging features, which when correlated with genomics, is termed radiogenomics. However, radiogenomic biological validation is not sufficiently described in the literature. We seek to establish causality between differential gene expression status and MRI-extracted radiomic-features in glioblastoma. Experimental Design: Radiogenomic predictions and validation were done using the Cancer Genome Atlas and Repository of Molecular Brain Neoplasia Data glioblastoma patients (n = 93) and orthotopic xenografts (OX; n = 40). Tumor phenotypes were segmented, and radiomic-features extracted using the developed radiome-sequencing pipeline. Patients and animals were dichotomized on the basis of Periostin (POSTN) expression levels. RNA and protein levels confirmed RNAi-mediated POSTN knockdown in OX. Total RNA of tumor cells isolated from mouse brains (knockdown and control) was used for microarray-based expression profiling. Radiomic-features were utilized to predict POSTN expression status in patient, mouse, and interspecies. Results: Our robust pipeline consists of segmentation, radiomic-feature extraction, feature normalization/selection, and predictive modeling. The combination of skull stripping, brain-tissue focused normalization, and patient-specific normalization are unique to this study, providing comparable cross-platform, cross-institution radiomic features. POSTN expression status was not associated with qualitative or volumetric MRI parameters. Radiomic features significantly predicted POSTN expression status in patients (AUC: 76.56%; sensitivity/specificity: 73.91/78.26%) and OX (AUC: 92.26%; sensitivity/specificity: 92.86%/91.67%). Furthermore, radiomic features in OX were significantly associated with patients with similar POSTN expression levels (AUC: 93.36%; sensitivity/specificity: 82.61%/95.74%; P = 02.021E−15). Conclusions: We determined causality between radiomic texture features and POSTN expression levels in a preclinical model with clinical validation. Our biologically validated radiomic pipeline also showed the potential application for human–mouse matched coclinical trials.

Silent Sentence Completion Shows Superiority Localizing Wernicke's Area and Activation Patterns of Distinct Language Paradigms Correlate with Genomics: Prospective Study

Preoperative mapping of language areas using fMRI greatly depends on the paradigms used, as different tasks harness distinct capabilities to activate speech processing areas. In this study, we compared the ability of 3 covert speech paradigms: Silent Sentence Completion (SSC), category naming (CAT) and verbal fluency (FAS), in localizing the Wernicke’s area and studied the association between genomic markers and functional activation. Fifteen right-handed healthy volunteers and 35 mixed-handed patients were included. We focused on the anatomical areas of posterosuperior, middle temporal and angular gyri corresponding to Wernicke’s area. Activity was deemed significant in a region of interest if P < 0.05. Association between fMRI activation and genomic mutation status was obtained. Results demonstrated SSC’s superiority at localizing Wernicke’s area. SSC demonstrated functional activity in 100% of cancer patients and healthy volunteers; which was significantly higher than those for FAS and CAT. Patients with 1p/19q non-co-deleted had higher extent of activation on SSC (P < 0.02). Those with IDH-1 wild-type were more likely to show no activity on CAT (P < 0.05). SSC is a robust paradigm for localizing Wernicke’s area, making it an important clinical tool for function-preserving surgeries. We also found a correlation between tumor genomics and functional activation, which deserves more comprehensive study.

Abstract 4217: First pre-clinical validation of radiogenomics in glioblastoma

A plethora of Magnetic Resonance Imaging (MRI) features have been correlated to cancer genomics to date, however, none have established causality. Here, we present an in vivo xenograft RNA interference validated, potentially clinically applicable test method termed “Magnetic Resonance Radiomic Sequencing” (MRRS) for the noninvasive detection of cancer genomics in Glioblastoma. MRRS comprehensively assesses the entire tumor mass using imaging texture-based algorithms that generate thousands of variables (features) inherent to the tumor. Two independent glioblastoma stem cells (GSC1 and GSC3) harboring doxycycline inducible short hairpin RNA against Periostin (POSTN), a gene previously identified in our radiogenomic screen, were implanted at orthotopic location in nude mouse brain. In vivo knockdown of >90% and ∼40% POSTN gene was achieved in GSC3 and GSC1 respectively. The T2 and T1 post MRI texture features, in edema and contrast enhancement phenotype features were compared between doxycycline (POSTN knockdown) and sucrose (control) group of mice using T test statistics. The significant features were included in a Stepwise Forward Logistic Regression analysis to build the final predictive model. The accuracy of the model was tested using ROC cure analysis. Among 3600 features in GSC3 mice cohort, 117 features were significantly (p value Citation Format: Pascal Zinn, Sanjay Singh, Markus M. Luedi, Faramak Zandi, Aikaterini Kotrotsou, Masumeh Hatami, Ginu Thomas, Ahmed Elakkad, Joy Gumin, Erik P. Sulman, Frederick Lang, David Piwnica-Worms, Rivka R. Colen. First pre-clinical validation of radiogenomics in glioblastoma. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 4217.