J. Ross Mitchell

Radiogenomics to characterize regional genetic heterogeneity in glioblastoma

Background Glioblastoma (GBM) exhibits profound intratumoral genetic heterogeneity. Each tumor comprises multiple genetically distinct clonal populations with different therapeutic sensitivities. This has implications for targeted therapy and genetically informed paradigms. Contrast-enhanced (CE)-MRI and conventional sampling techniques have failed to resolve this heterogeneity, particularly for nonenhancing tumor populations. This study explores the feasibility of using multiparametric MRI and texture analysis to characterize regional genetic heterogeneity throughout MRI-enhancing and nonenhancing tumor segments. Methods We collected multiple image-guided biopsies from primary GBM patients throughout regions of enhancement (ENH) and nonenhancing parenchyma (so called brain-around-tumor, [BAT]). For each biopsy, we analyzed DNA copy number variants for core GBM driver genes reported by The Cancer Genome Atlas. We co-registered biopsy locations with MRI and texture maps to correlate regional genetic status with spatially matched imaging measurements. We also built multivariate predictive decision-tree models for each GBM driver gene and validated accuracies using leave-one-out-cross-validation (LOOCV). Results We collected 48 biopsies (13 tumors) and identified significant imaging correlations (univariate analysis) for 6 driver genes: EGFR, PDGFRA, PTEN, CDKN2A, RB1, and TP53. Predictive model accuracies (on LOOCV) varied by driver gene of interest. Highest accuracies were observed for PDGFRA (77.1%), EGFR (75%), CDKN2A (87.5%), and RB1 (87.5%), while lowest accuracy was observed in TP53 (37.5%). Models for 4 driver genes (EGFR, RB1, CDKN2A, and PTEN) showed higher accuracy in BAT samples (n = 16) compared with those from ENH segments (n = 32). Conclusion MRI and texture analysis can help characterize regional genetic heterogeneity, which offers potential diagnostic value under the paradigm of individualized oncology.

Multi-Parametric MRI and Texture Analysis to Visualize Spatial Histologic Heterogeneity and Tumor Extent in Glioblastoma

Background Genetic profiling represents the future of neuro-oncology but suffers from inadequate biopsies in heterogeneous tumors like Glioblastoma (GBM). Contrast-enhanced MRI (CE-MRI) targets enhancing core (ENH) but yields adequate tumor in only ~60% of cases. Further, CE-MRI poorly localizes infiltrative tumor within surrounding non-enhancing parenchyma, or brain-around-tumor (BAT), despite the importance of characterizing this tumor segment, which universally recurs. In this study, we use multiple texture analysis and machine learning (ML) algorithms to analyze multi-parametric MRI, and produce new images indicating tumor-rich targets in GBM. Methods We recruited primary GBM patients undergoing image-guided biopsies and acquired pre-operative MRI: CE-MRI, Dynamic-Susceptibility-weighted-Contrast-enhanced-MRI, and Diffusion Tensor Imaging. Following image coregistration and region of interest placement at biopsy locations, we compared MRI metrics and regional texture with histologic diagnoses of high- vs low-tumor content (≥80% vs <80% tumor nuclei) for corresponding samples. In a training set, we used three texture analysis algorithms and three ML methods to identify MRI-texture features that optimized model accuracy to distinguish tumor content. We confirmed model accuracy in a separate validation set. Results We collected 82 biopsies from 18 GBMs throughout ENH and BAT. The MRI-based model achieved 85% cross-validated accuracy to diagnose high- vs low-tumor in the training set (60 biopsies, 11 patients). The model achieved 81.8% accuracy in the validation set (22 biopsies, 7 patients). Conclusion Multi-parametric MRI and texture analysis can help characterize and visualize GBM’s spatial histologic heterogeneity to identify regional tumor-rich biopsy targets.

Validation study of a fast, accurate, and precise brain tumor volume measurement

UNLABELLED Precision and accuracy are sometimes sacrificed to ensure that medical image processing is rapid. To address this, our lab had developed a novel level set segmentation algorithm that is 16× faster and >96% accurate on realistic brain phantoms. METHODS This study reports speed, precision and estimated accuracy of our algorithm when measuring MRIs of meningioma brain tumors and compares it to manual tracing and modified MacDonald (MM) ellipsoid criteria. A repeated-measures study allowed us to determine measurement precisions (MPs) - clinically relevant thresholds for statistically significant change. RESULTS Speed: the level set, MM, and trace methods required 1:20, 1:35, and 9:35 (mm:ss) respectively on average to complete a volume measurement (p<0.05). Accuracy: the level set was not statistically different to the estimated true lesion volumes (p>0.05). Precision: the MMs within-operator and between-operator MPs were significantly higher (worse) than the other methods (p<0.05). The observed difference in MP between the level set and trace methods did not reach statistical significance (p>0.05). CONCLUSION Our level set is faster on average than MM, yet has accuracy and precision comparable to manual tracing.

Computer-aided diagnosis of contrast-enhanced spectral mammography: A feasibility study

OBJECTIVE To evaluate whether the use of a computer-aided diagnosis-contrast-enhanced spectral mammography (CAD-CESM) tool can further increase the diagnostic performance of CESM compared with that of experienced radiologists. MATERIALS AND METHODS This IRB-approved retrospective study analyzed 50 lesions described on CESM from August 2014 to December 2015. Histopathologic analyses, used as the criterion standard, revealed 24 benign and 26 malignant lesions. An expert breast radiologist manually outlined lesion boundaries on the different views. A set of morphologic and textural features were then extracted from the low-energy and recombined images. Machine-learning algorithms with feature selection were used along with statistical analysis to reduce, select, and combine features. Selected features were then used to construct a predictive model using a support vector machine (SVM) classification method in a leave-one-out-cross-validation approach. The classification performance was compared against the diagnostic predictions of 2 breast radiologists with access to the same CESM cases. RESULTS Based on the SVM classification, CAD-CESM correctly identified 45 of 50 lesions in the cohort, resulting in an overall accuracy of 90%. The detection rate for the malignant group was 88% (3 false-negative cases) and 92% for the benign group (2 false-positive cases). Compared with the model, radiologist 1 had an overall accuracy of 78% and a detection rate of 92% (2 false-negative cases) for the malignant group and 62% (10 false-positive cases) for the benign group. Radiologist 2 had an overall accuracy of 86% and a detection rate of 100% for the malignant group and 71% (8 false-positive cases) for the benign group. CONCLUSIONS The results of our feasibility study suggest that a CAD-CESM tool can provide complementary information to radiologists, mainly by reducing the number of false-positive findings.

Active inflammation increases the heterogeneity of MRI texture in mice with relapsing experimental allergic encephalomyelitis.

Inflammation modulates tissue damage in relapsing-remitting multiple sclerosis (MS) both acutely and chronically, but its severity is difficult to evaluate with conventional MRI analysis. In mice with experimental allergic encephalomyelitis (EAE, a model of MS), we administered ultra small particles of iron oxide to track macrophage-mediated inflammation during the onset (relapse) and recovery (remission) of disease activity using high field MRI. We performed MRI texture analysis, a sensitive measure of tissue regularity, and T2 assessment both in EAE lesions and the control tissue, and measured spinal cord volume. We found that inflammation was 3 times more remarkable at onset than at recovery of EAE in histology yet demyelination appeared similar across animals and disease course. In MRI, lesion texture was more heterogeneous; T2 was lower; and spinal cord volume was greater in EAE than in controls, but only MRI texture was worse at relapse than at remission of EAE. Moreover, MRI texture correlated with spinal cord volume and tended to correlate with the extent of disability in EAE. While subject to further confirmation, our findings may suggest the sensitivity of MRI texture analysis for accessing inflammation.

Pilot study: Evaluation of dual-energy computed tomography measurement strategies for positron emission tomography correlation in pancreatic adenocarcinoma.

We sought to determine whether dual-energy computed tomography (DECT) measurements correlate with positron emission tomography (PET) standardized uptake values (SUVs) in pancreatic adenocarcinoma, and to determine the optimal DECT imaging variables and modeling strategy to produce the highest correlation with maximum SUV (SUVmax). We reviewed 25 patients with unresectable pancreatic adenocarcinoma seen at Mayo Clinic, Scottsdale, Arizona, who had PET–computed tomography (PET/CT) and enhanced DECT performed the same week between March 25, 2010 and December 9, 2011. For each examination, DECT measurements were taken using one of three methods: (1) average values of three tumor regions of interest (ROIs) (method 1); (2) one ROI in the area of highest subjective DECT enhancement (method 2); and (3) one ROI in the area corresponding to PET SUVmax (method 3). There were 133 DECT variables using method 1, and 89 using the other methods. Univariate and multivariate analysis regression models were used to identify important correlations between DECT variables and PET SUVmax. Both R2 and adjusted R2 were calculated for the multivariate model to compensate for the increased number of predictors. The average SUVmax was 5 (range, 1.8–12.0). Multivariate analysis of DECT imaging variables outperformed univariate analysis (r = 0.91; R2 = 0.82; adjusted R2 = 0.75 vs r < 0.58; adjusted R2 < 0.34). Method 3 had the highest correlation with PET SUVmax (R2 = 0.82), followed by method 1 (R2 = 0.79) and method 2 (R2 = 0.57). DECT thus has clinical potential as a surrogate for, or as a complement to, PET in patients with pancreatic adenocarcinoma.

Computed Tomography-Based Texture Analysis to Determine Human Papillomavirus Status of Oropharyngeal Squamous Cell Carcinoma

Objective To determine whether machine learning can accurately classify human papillomavirus (HPV) status of oropharyngeal squamous cell carcinoma (OPSCC) using computed tomography (CT)-based texture analysis. Methods Texture analyses were retrospectively applied to regions of interest from OPSCC primary tumors on contrast-enhanced neck CT, and machine learning was used to create a model that classified HPV status with the highest accuracy. Results were compared against the blinded review of 2 neuroradiologists. Results The HPV-positive (n = 92) and -negative (n = 15) cohorts were well matched clinically. Neuroradiologist classification accuracies for HPV status (44.9%, 55.1%) were not significantly different (P = 0.13), and there was a lack of agreement between the 2 neuroradiologists (&kgr; = −0.145). The best machine learning model had an accuracy of 75.7%, which was greater than either neuroradiologist (P < 0.001, P = 0.002). Conclusions Useful diagnostic information regarding HPV infection can be extracted from the CT appearance of OPSCC beyond what is apparent to the trained human eye.

Abstract A08: Histologic evidence for a bio-mathematical model of glioblastoma invasion

Purpose: Investigate the utility of patient-specific spatial predictions of tumor cell density from a bio-mathematical model. Introduction: Glioblastomas (GBMs) are the most malignant of all primary brain tumors. While it is known there is always a non-detectable portion of the tumor, current techniques of monitoring GBM progression, imaging and initial histological assessment, are not able to reliably estimate the tumor invasion past the enhancing region on T2-Weighted (T2W) imaging. Over the last two decades, a large effort has been made to create a simple patient-specific mathematical model of gliomas. The resulting model, referred to as the Proliferation-Invasion (PI) model, is based on two key parameters, the net growth rate, ρ, and the dispersal coefficient, D. In this model, the ratio of D/ρ is related to degree of invasion and the product D*ρ, is related to the speed of growth. The intuitive understanding provided by this model has been able to provide patient-specific understanding of disease kinetics enabling prediction of outcomes following surgical resection, radiation and the development of a prognostic response metric. Previous literature utilizing this model has been based on the assumption that what is seen on the pretreatment T1-Weighted contrast-enhanced (T1Gd) and T2W, images correspond to an 80% and 16% tumor cell density threshold respectively. This assumption allows for an estimate of D/ρ from a single time point of imaging. While these values were based on extensive experience, for ethical and technical reasons, they have never been rigorously investigated histologically. Recent technological advances have made it possible for surgeons to use an MRI to guide the acquisition of tissue making it possible to know with a good degree of accuracy where on the MR image the histological specimen comes from. Methods: Model Calibration : To estimate D/ρ for each patient, we assume abnormalities on the T1Gd and T2W images correspond to an 80% and 16% tumor cell density threshold respectively. We then utilize a Bayesian calibration approach based on adaptive grid refinement while holding the velocity constant to find the most likely value of D/ρ to match the observed radial measurements. Three-Dimensional Density Maps : Given a gray/white segmentation and an estimate for D/ρ, we can build a tumor cell density prediction in the patient9s anatomy using the Eikonal equations and the modified Fast Marching Method (FMM) algorithm presented by Konukoglu et al. Patient Cohort : Eighteen patients were recruited with clinically suspected GBM undergoing preoperative stereotactic MRI for surgical resection with IRB approval Barrow Neurological Institute and Mayo Clinic in Arizona. Surgical Biopsy : Pre-operative conventional MRI, including T1Gd and T2W, was utilized to guide stereotactic biopsies. An average of 5–6 tissue specimens were acquired from each tumor by using stereotactic surgical localization, following the smallest possible diameter craniotomies to minimize brain shift. Histological Analysis : 4 μm tissue sections were stained with hematoxylin and eosin (HE 2016 Jun 25-28; Boston, MA. Philadelphia (PA): AACR; Cancer Res 2017;77(2 Suppl):Abstract nr A08.

An Introduction to Radiomics: An Evolving Cornerstone of Precision Medicine

Abstract Over the past decade a compelling body of literature has emerged suggesting a more pivotal role for imaging in the diagnosis, prognosis, and monitoring of diseases. These advances have facilitated the rise of an emerging practice known as Radiomics: the extraction and analysis of large numbers of quantitative features from medical images to improve disease characterization and prediction of outcome. Our aim is to clarify the potential clinical utility of radiomics specifically in solid tumor cancer care. First, we discuss the clinical processes of cancer management (i.e., diagnosis, monitoring, and treatment). Second, the limitations of the current process are discussed followed by a discourse of the benefits of radiomics. Third, we provide a high-level description of the workflow of radiomics and explain how it relates to biomedical texture analysis. We conclude with a review of the literature in this field and the remaining challenges ahead.

Pilot Study: Two-Stage Hybrid Model to Correlate Single Energy CT and PET in Pancreatic Adenocarcinoma

Pancreas adenocarcinoma is one of the most common malignant tumors and the fourth leading cause of cancerrelated mortality. While Computed Tomography (CT) has been commonly used clinically for the cancer staging and follow-up, Positron Emission Tomography (PET) is known to be generally more accurate and sensitive for metastases and thus has great prognostic value. However, PET is more expensive and less accessible. This research is to explore the use of multivariate models to extract valuable information from CT to mimic the effects of PET. Based on the original 6 CT measures, 10 CT biomarkers are derived. The strongest correlation with PET SUV in the multivariate regression on the 6 original measures is r2=0.41 (r=0.64), on the 10 derived biomarkers is r2=0.55 (r=0.74). We developed a twostage hybrid model, where a multivariate classifier was developed to first separate the patients into the group with high SUV values vs. low SUV values, then the regression model was developed for each group respectively. The overall performance of this two-stage model is more promising with an r2=0.81 (r=0.90). We conclude advanced CT analytics has the potential to extract valuable information that correlates with PET SUV. Rationale and objectives: Pancreatic adenocarcinoma is commonly studied by CT and PET. We aimed to see if information from CT could be used to simulate the results of PET. Materials and methods: A retrospective study of 24 patients with pancreatic cancer who had both CT and PET in close temporal proximity was conducted. Measurements of the aorta, normal pancreatic tissue, solid and cystic portions of pancreatic tumors were performed resulting in 6 biomarkers. Ten more biomarkers were derived including the ratios of solid and cystic tumor mean and standard deviation to normal pancreas (and to each other), as well as signal-to-noise ratios of solid and cystic tumors to normal pancreas. Univariate analysis and multivariate regression were conducted on the original measures (6 biomarkers) and derived measures (10 biomarkers). A two-stage hybrid model integrating machine learning model with multivariate regression analysis was also studied. Results: The best results were obtained using the two-stage hybrid model. The regression model for low SUV (≤5) used cystic tumor mean (r2=0.68, r=0.83). The regression model for high SUV(>5) used tumor mean, the ratios of tumor mean to pancreas mean, tumor mean to aorta mean, standard deviation of tumor to aorta mean and signal-to-noise ratio of difference between the normal pancreas mean and solid tumor mean to standard deviation of pancreas (r2=0.86, r=0.93). The overall performance of the two-stage model is r2=0.81(r=0.90). Conclusion: Two-stage multivariate analysis of CT parameters can mimic the effects of PET to a reasonable extent, and signal-to-noise and standard deviation ratios may capture the essential nonlinearity of these relationships.