An Automated Pipeline for Differential Cell Counts on Whole-Slide Bone Marrow Aspirate Smears.

The pathologic diagnosis of bone marrow disorders relies in part on the microscopic analysis of bone marrow aspirate (BMA) smears and the manual counting of marrow nucleated cells to obtain a differential cell count (DCC). This manual process has significant limitations, including the analysis of only a small subset of optimal slide areas and nucleated cells, as well as interobserver variability due to differences in cell selection and classification. To address these shortcomings, we developed an automated machine learning-based pipeline for obtaining 11-component DCCs on whole-slide BMAs. This pipeline uses a sequential process of identifying optimal BMA regions with high proportions of marrow nucleated cells, detecting individual cells within these optimal areas, and classifying these cells into 1 of 11 DCC components. Convolutional neural network models were trained on 396,048 BMA region, 28,914 cell boundary, and 1,510,976 cell class images from manual annotations. The resulting automated pipeline produced 11-component DCCs that demonstrated a high statistical and diagnostic concordance with manual DCCs among a heterogeneous group of testing BMA slides with varying pathologies and cellularities. Additionally, we demonstrated that an automated analysis can reduce the intraslide variance in DCCs by analyzing the whole slide and marrow nucleated cells within all optimal regions. Finally, the pipeline outputs of region classification, cell detection, and cell classification can be visualized using whole-slide image analysis software. This study demonstrates the feasibility of a fully automated pipeline for generating DCCs on scanned whole-slide BMA images, with the potential for improving the current standard of practice for utilizing BMA smears in the laboratory analysis of hematologic disorders.

Explainable nucleus classification using Decision Tree Approximation of Learned Embeddings.

MOTIVATION: Nucleus detection, segmentation and classification are fundamental to high-resolution mapping of the tumor microenvironment using whole-slide histopathology images. The growing interest in leveraging the power of deep learning to achieve state-of-the-art performance often comes at the cost of explainability, yet there is general consensus that explainability is critical for trustworthiness and widespread clinical adoption. Unfortunately, current explainability paradigms that rely on pixel saliency heatmaps or superpixel importance scores are not well-suited for nucleus classification. Techniques like Grad-CAM or LIME provide explanations that are indirect, qualitative and/or nonintuitive to pathologists. RESULTS: In this article, we present techniques to enable scalable nuclear detection, segmentation and explainable classification. First, we show how modifications to the widely used Mask R-CNN architecture, including decoupling the detection and classification tasks, improves accuracy and enables learning from hybrid annotation datasets like NuCLS, which contain mixtures of bounding boxes and segmentation boundaries. Second, we introduce an explainability method called Decision Tree Approximation of Learned Embeddings (DTALE), which provides explanations for classification model behavior globally, as well as for individual nuclear predictions. DTALE explanations are simple, quantitative, and can flexibly use any measurable morphological features that make sense to practicing pathologists, without sacrificing model accuracy. Together, these techniques present a step toward realizing the promise of computational pathology in computer-aided diagnosis and discovery of morphologic biomarkers. AVAILABILITY AND IMPLEMENTATION: Relevant code can be found at github.com/CancerDataScience/NuCLS. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Neuromorphological changes following selection for tameness and aggression in the Russian fox-farm experiment.

The Russian fox-farm experiment is an unusually long-running and well-controlled study designed to replicate wolf-to-dog domestication. As such, it offers an unprecedented window onto the neural mechanisms governing the evolution of behavior. Here we report evolved changes to gray matter morphology resulting from selection for tameness vs. aggressive responses toward humans in a sample of 30 male fox brains. Contrasting with standing ideas on the effects of domestication on brain size, tame foxes did not show reduced brain volume. Rather, gray matter volume in both the tame and aggressive strains was increased relative to conventional farm foxes bred without deliberate selection on behavior. Furthermore, tame- and aggressive-enlarged regions overlapped substantially, including portions of motor, somatosensory, and prefrontal cortex, amygdala, hippocampus, and cerebellum. We also observed differential morphological covariation across distributed gray matter networks. In one prefrontal-cerebellum network, this covariation differentiated the three populations along the tame-aggressive behavioral axis. Surprisingly, a prefrontal-hypothalamic network differentiated the tame and aggressive foxes together from the conventional strain. These findings indicate that selection for opposite behaviors can influence brain morphology in a similar way.SIGNIFICANCE STATEMENTDomestication represents one of the largest and most rapid evolutionary shifts of life on earth. However, its neural correlates are largely unknown. Here we report the neuroanatomical consequences of selective breeding for tameness or aggression in the seminal Russian fox-farm experiment. Compared to a population of conventional farm-bred control foxes, tame foxes show neuroanatomical changes in the prefrontal cortex and hypothalamus, paralleling wolf-to-dog shifts. Surprisingly, though, aggressive foxes also show similar changes. Moreover, both strains show increased gray matter volume relative to controls. These results indicate that similar brain adaptations can result from selection for opposite behavior, that existing ideas of brain changes in domestication may need revision, and that significant neuroanatomical change can evolve very quickly - within the span of less than a hundred generations.

Predicting cancer outcomes from histology and genomics using convolutional networks.

Cancer histology reflects underlying molecular processes and disease progression and contains rich phenotypic information that is predictive of patient outcomes. In this study, we show a computational approach for learning patient outcomes from digital pathology images using deep learning to combine the power of adaptive machine learning algorithms with traditional survival models. We illustrate how these survival convolutional neural networks (SCNNs) can integrate information from both histology images and genomic biomarkers into a single unified framework to predict time-to-event outcomes and show prediction accuracy that surpasses the current clinical paradigm for predicting the overall survival of patients diagnosed with glioma. We use statistical sampling techniques to address challenges in learning survival from histology images, including tumor heterogeneity and the need for large training cohorts. We also provide insights into the prediction mechanisms of SCNNs, using heat map visualization to show that SCNNs recognize important structures, like microvascular proliferation, that are related to prognosis and that are used by pathologists in grading. These results highlight the emerging role of deep learning in precision medicine and suggest an expanding utility for computational analysis of histology in the future practice of pathology.

Significant Neuroanatomical Variation Among Domestic Dog Breeds.

Humans have bred different lineages of domestic dogs for different tasks such as hunting, herding, guarding, or companionship. These behavioral differences must be the result of underlying neural differences, but surprisingly, this topic has gone largely unexplored. The current study examined whether and how selective breeding by humans has altered the gross organization of the brain in dogs. We assessed regional volumetric variation in MRI studies of 62 male and female dogs of 33 breeds. Neuroanatomical variation is plainly visible across breeds. This variation is distributed nonrandomly across the brain. A whole-brain, data-driven independent components analysis established that specific regional subnetworks covary significantly with each other. Variation in these networks is not simply the result of variation in total brain size, total body size, or skull shape. Furthermore, the anatomy of these networks correlates significantly with different behavioral specialization(s) such as sight hunting, scent hunting, guarding, and companionship. Importantly, a phylogenetic analysis revealed that most change has occurred in the terminal branches of the dog phylogenetic tree, indicating strong, recent selection in individual breeds. Together, these results establish that brain anatomy varies significantly in dogs, likely due to human-applied selection for behavior.SIGNIFICANCE STATEMENT Dog breeds are known to vary in cognition, temperament, and behavior, but the neural origins of this variation are unknown. In an MRI-based analysis, we found that brain anatomy covaries significantly with behavioral specializations such as sight hunting, scent hunting, guarding, and companionship. Neuroanatomical variation is not simply driven by brain size, body size, or skull shape, and is focused in specific networks of regions. Nearly all of the identified variation occurs in the terminal branches of the dog phylogenetic tree, indicating strong, recent selection in individual breeds. These results indicate that through selective breeding, humans have significantly altered the brains of different lineages of domestic dogs in different ways.

Machine-based detection and classification for bone marrow aspirate differential counts: initial development focusing on nonneoplastic cells.

Bone marrow aspirate (BMA) differential cell counts (DCCs) are critical for the classification of hematologic disorders. While manual counts are considered the gold standard, they are labor intensive, time consuming, and subject to bias. A reliable automated counter has yet to be developed, largely due to the inherent complexity of bone marrow specimens. Digital pathology imaging coupled with machine learning algorithms represents a highly promising emerging technology for this purpose. Yet, training datasets for BMA cellular constituents, critical for building and validating machine learning algorithms, are lacking. Herein, we report our experience creating and employing such datasets to develop a machine learning algorithm to detect and classify BMA cells. Utilizing a web-based system that we developed for annotating and managing digital pathology images, over 10,000 cells from scanned whole slide images of BMA smears were manually annotated, including all classes that comprise the standard clinical DCC. We implemented a two-stage, detection and classification approach that allows design flexibility and improved classification accuracy. In a sixfold cross-validation, our algorithms achieved high overall accuracy in detection (0.959 ± 0.008 precision-recall AUC) and classification (0.982 ± 0.03 ROC AUC) using nonneoplastic samples. Testing on a small set of acute myeloid leukemia and multiple myeloma samples demonstrated similar detection and classification performance. In summary, our algorithms showed promising early results and represent an important initial step in the effort to devise a reliable, objective method to automate DCCs. With further development to include formal clinical validation, such a system has the potential to assist in disease diagnosis and prognosis, and significantly impact clinical practice.

Structured crowdsourcing enables convolutional segmentation of histology images.

MOTIVATION: While deep-learning algorithms have demonstrated outstanding performance in semantic image segmentation tasks, large annotation datasets are needed to create accurate models. Annotation of histology images is challenging due to the effort and experience required to carefully delineate tissue structures, and difficulties related to sharing and markup of whole-slide images. RESULTS: We recruited 25 participants, ranging in experience from senior pathologists to medical students, to delineate tissue regions in 151 breast cancer slides using the Digital Slide Archive. Inter-participant discordance was systematically evaluated, revealing low discordance for tumor and stroma, and higher discordance for more subjectively defined or rare tissue classes. Feedback provided by senior participants enabled the generation and curation of 20 000+ annotated tissue regions. Fully convolutional networks trained using these annotations were highly accurate (mean AUC=0.945), and the scale of annotation data provided notable improvements in image classification accuracy. AVAILABILITY AND IMPLEMENTATION: Dataset is freely available at: https://goo.gl/cNM4EL. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Computer algorithms show potential for improving dermatologists' accuracy to diagnose cutaneous melanoma: Results of the International Skin Imaging Collaboration 2017.

BACKGROUND: Computer vision has promise in image-based cutaneous melanoma diagnosis but clinical utility is uncertain. OBJECTIVE: To determine if computer algorithms from an international melanoma detection challenge can improve dermatologists' accuracy in diagnosing melanoma. METHODS: In this cross-sectional study, we used 150 dermoscopy images (50 melanomas, 50 nevi, 50 seborrheic keratoses) from the test dataset of a melanoma detection challenge, along with algorithm results from 23 teams. Eight dermatologists and 9 dermatology residents classified dermoscopic lesion images in an online reader study and provided their confidence level. RESULTS: The top-ranked computer algorithm had an area under the receiver operating characteristic curve of 0.87, which was higher than that of the dermatologists (0.74) and residents (0.66) (P < .001 for all comparisons). At the dermatologists' overall sensitivity in classification of 76.0%, the algorithm had a superior specificity (85.0% vs. 72.6%, P = .001). Imputation of computer algorithm classifications into dermatologist evaluations with low confidence ratings (26.6% of evaluations) increased dermatologist sensitivity from 76.0% to 80.8% and specificity from 72.6% to 72.8%. LIMITATIONS: Artificial study setting lacking the full spectrum of skin lesions as well as clinical metadata. CONCLUSION: Accumulating evidence suggests that deep neural networks can classify skin images of melanoma and its benign mimickers with high accuracy and potentially improve human performance.

The molecular basis of breast cancer pathological phenotypes.

The histopathological evaluation of morphological features in breast tumours provides prognostic information to guide therapy. Adjunct molecular analyses provide further diagnostic, prognostic and predictive information. However, there is limited knowledge of the molecular basis of morphological phenotypes in invasive breast cancer. This study integrated genomic, transcriptomic and protein data to provide a comprehensive molecular profiling of morphological features in breast cancer. Fifteen pathologists assessed 850 invasive breast cancer cases from The Cancer Genome Atlas (TCGA). Morphological features were significantly associated with genomic alteration, DNA methylation subtype, PAM50 and microRNA subtypes, proliferation scores, gene expression and/or reverse-phase protein assay subtype. Marked nuclear pleomorphism, necrosis, inflammation and a high mitotic count were associated with the basal-like subtype, and had a similar molecular basis. Omics-based signatures were constructed to predict morphological features. The association of morphology transcriptome signatures with overall survival in oestrogen receptor (ER)-positive and ER-negative breast cancer was first assessed by use of the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) dataset; signatures that remained prognostic in the METABRIC multivariate analysis were further evaluated in five additional datasets. The transcriptomic signature of poorly differentiated epithelial tubules was prognostic in ER-positive breast cancer. No signature was prognostic in ER-negative breast cancer. This study provided new insights into the molecular basis of breast cancer morphological phenotypes. The integration of morphological with molecular data has the potential to refine breast cancer classification, predict response to therapy, enhance our understanding of breast cancer biology, and improve clinical management. This work is publicly accessible at www.dx.ai/tcga_breast. Copyright © 2016 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

Results of the 2016 International Skin Imaging Collaboration International Symposium on Biomedical Imaging challenge: Comparison of the accuracy of computer algorithms to dermatologists for the diagnosis of melanoma from dermoscopic images.

BACKGROUND: Computer vision may aid in melanoma detection. OBJECTIVE: We sought to compare melanoma diagnostic accuracy of computer algorithms to dermatologists using dermoscopic images. METHODS: We conducted a cross-sectional study using 100 randomly selected dermoscopic images (50 melanomas, 44 nevi, and 6 lentigines) from an international computer vision melanoma challenge dataset (n = 379), along with individual algorithm results from 25 teams. We used 5 methods (nonlearned and machine learning) to combine individual automated predictions into "fusion" algorithms. In a companion study, 8 dermatologists classified the lesions in the 100 images as either benign or malignant. RESULTS: The average sensitivity and specificity of dermatologists in classification was 82% and 59%. At 82% sensitivity, dermatologist specificity was similar to the top challenge algorithm (59% vs. 62%, P = .68) but lower than the best-performing fusion algorithm (59% vs. 76%, P = .02). Receiver operating characteristic area of the top fusion algorithm was greater than the mean receiver operating characteristic area of dermatologists (0.86 vs. 0.71, P = .001). LIMITATIONS: The dataset lacked the full spectrum of skin lesions encountered in clinical practice, particularly banal lesions. Readers and algorithms were not provided clinical data (eg, age or lesion history/symptoms). Results obtained using our study design cannot be extrapolated to clinical practice. CONCLUSION: Deep learning computer vision systems classified melanoma dermoscopy images with accuracy that exceeded some but not all dermatologists.

Gene integrated set profile analysis: a context-based approach for inferring biological endpoints.

The identification of genes with specific patterns of change (e.g. down-regulated and methylated) as phenotype drivers or samples with similar profiles for a given gene set as drivers of clinical outcome, requires the integration of several genomic data types for which an 'integrate by intersection' (IBI) approach is often applied. In this approach, results from separate analyses of each data type are intersected, which has the limitation of a smaller intersection with more data types. We introduce a new method, GISPA (Gene Integrated Set Profile Analysis) for integrated genomic analysis and its variation, SISPA (Sample Integrated Set Profile Analysis) for defining respective genes and samples with the context of similar, a priori specified molecular profiles. With GISPA, the user defines a molecular profile that is compared among several classes and obtains ranked gene sets that satisfy the profile as drivers of each class. With SISPA, the user defines a gene set that satisfies a profile and obtains sample groups of profile activity. Our results from applying GISPA to human multiple myeloma (MM) cell lines contained genes of known profiles and importance, along with several novel targets, and their further SISPA application to MM coMMpass trial data showed clinical relevance.

PACAP receptor gene polymorphism impacts fear responses in the amygdala and hippocampus.

We have recently found higher circulating levels of pituitary adenylate cyclase-activating polypeptide (PACAP) associated with posttraumatic stress disorder (PTSD) symptoms in a highly traumatized cohort of women but not men. Furthermore, a single nucleotide polymorphism in the PACAP receptor gene ADCYAP1R1, adenylate cyclase activating polypeptide 1 receptor type 1, was associated with individual differences in PTSD symptoms and psychophysiological markers of fear and anxiety. The current study outlines an investigation of individual differences in brain function associated with ADCYAP1R1 genotype. Forty-nine women who had experienced moderate to high levels of lifetime trauma participated in a functional MRI task involving passive viewing of threatening and neutral face stimuli. Analyses focused on the amygdala and hippocampus, regions that play central roles in the pathophysiology of PTSD and are known to have high densities of PACAP receptors. The risk genotype was associated with increased reactivity of the amygdala and hippocampus to threat stimuli and decreased functional connectivity between the amygdala and hippocampus. The findings indicate that the PACAP system modulates medial temporal lobe function in humans. Individual differences in ADCYAP1R1 genotype may contribute to dysregulated fear circuitry known to play a central role in PTSD and other anxiety disorders.

Imaging-genomics reveals driving pathways of MRI derived volumetric tumor phenotype features in Glioblastoma.

BACKGROUND: Glioblastoma (GBM) tumors exhibit strong phenotypic differences that can be quantified using magnetic resonance imaging (MRI), but the underlying biological drivers of these imaging phenotypes remain largely unknown. An Imaging-Genomics analysis was performed to reveal the mechanistic associations between MRI derived quantitative volumetric tumor phenotype features and molecular pathways. METHODS: One hundred fourty one patients with presurgery MRI and survival data were included in our analysis. Volumetric features were defined, including the necrotic core (NE), contrast-enhancement (CE), abnormal tumor volume assessed by post-contrast T1w (tumor bulk or TB), tumor-associated edema based on T2-FLAIR (ED), and total tumor volume (TV), as well as ratios of these tumor components. Based on gene expression where available (n = 91), pathway associations were assessed using a preranked gene set enrichment analysis. These results were put into context of molecular subtypes in GBM and prognostication. RESULTS: Volumetric features were significantly associated with diverse sets of biological processes (FDR < 0.05). While NE and TB were enriched for immune response pathways and apoptosis, CE was associated with signal transduction and protein folding processes. ED was mainly enriched for homeostasis and cell cycling pathways. ED was also the strongest predictor of molecular GBM subtypes (AUC = 0.61). CE was the strongest predictor of overall survival (C-index = 0.6; Noether test, p = 4x10(-4)). CONCLUSION: GBM volumetric features extracted from MRI are significantly enriched for information about the biological state of a tumor that impacts patient outcomes. Clinical decision-support systems could exploit this information to develop personalized treatment strategies on the basis of noninvasive imaging.

Applicability of digital analysis and imaging technology in neuropathology assessment.

Alzheimer's disease (AD) is a progressive neurological disorder that affects more than 30 million people worldwide. While various dementia-related losses in cognitive functioning are its hallmark clinical symptoms, ultimate diagnosis is based on manual neuropathological assessments using various schemas, including Braak staging, CERAD (Consortium to Establish a Registry for Alzheimer's Disease) and Thal phase scoring. Since these scoring systems are based on subjective assessment, there is inevitably some degree of variation between readers, which could affect ultimate neuropathology diagnosis. Here, we report a pilot study investigating the applicability of computer-driven image analysis for characterizing neuropathological features, as well as its potential to supplement or even replace manually derived ratings commonly performed in medical settings. In this work, we quantitatively measured amyloid beta (Aß) plaque in various brain regions from 34 patients using a robust digital quantification algorithm. We next verified these digitally derived measures to the manually derived pathology ratings using correlation and ordinal logistic regression methods, while also investigating the association with other AD-related neuropathology scoring schema commonly used at autopsy, such as Braak and CERAD. In addition to successfully verifying our digital measurements of Aß plaques with respective categorical measurements, we found significant correlations with most AD-related scoring schemas. Our results demonstrate the potential for digital analysis to be adapted to more complex staining procedures commonly used in neuropathological diagnosis. As the efficiency of scanning and digital analysis of histology images increases, we believe that the basis of our semi-automatic approach may better standardize quantification of neuropathological changes and AD diagnosis, ultimately leading to a more comprehensive understanding of neurological disorders and more efficient patient care.

A comparative analysis of mouse and human medial geniculate nucleus connectivity: a DTI and anterograde tracing study.

Understanding the function and connectivity of thalamic nuclei is critical for understanding normal and pathological brain function. The medial geniculate nucleus (MGN) has been studied mostly in the context of auditory processing and its connection to the auditory cortex. However, there is a growing body of evidence that the MGN and surrounding associated areas ('MGN/S') have a diversity of projections including those to the globus pallidus, caudate/putamen, amygdala, hypothalamus, and thalamus. Concomitantly, pathways projecting to the medial geniculate include not only the inferior colliculus but also the auditory cortex, insula, cerebellum, and globus pallidus. Here we expand our understanding of the connectivity of the MGN/S by using comparative diffusion weighted imaging with probabilistic tractography in both human and mouse brains (most previous work was in rats). In doing so, we provide the first report that attempts to match probabilistic tractography results between human and mice. Additionally, we provide anterograde tracing results for the mouse brain, which corroborate the probabilistic tractography findings. Overall, the study provides evidence for the homology of MGN/S patterns of connectivity across species for understanding translational approaches to thalamic connectivity and function. Further, it points to the utility of DTI in both human studies and small animal modeling, and it suggests potential roles of these connections in human cognition, behavior, and disease.

Virtual dissection and comparative connectivity of the superior longitudinal fasciculus in chimpanzees and humans.

Many of the behavioral capacities that distinguish humans from other primates rely on fronto-parietal circuits. The superior longitudinal fasciculus (SLF) is the primary white matter tract connecting lateral frontal with lateral parietal regions; it is distinct from the arcuate fasciculus, which interconnects the frontal and temporal lobes. Here we report a direct, quantitative comparison of SLF connectivity using virtual in vivo dissection of the SLF in chimpanzees and humans. SLF I, the superior-most branch of the SLF, showed similar patterns of connectivity between humans and chimpanzees, and was proportionally volumetrically larger in chimpanzees. SLF II, the middle branch, and SLF III, the inferior-most branch, showed species differences in frontal connectivity. In humans, SLF II showed greater connectivity with dorsolateral prefrontal cortex, whereas in chimps SLF II showed greater connectivity with the inferior frontal gyrus. SLF III was right-lateralized and proportionally volumetrically larger in humans, and human SLF III showed relatively reduced connectivity with dorsal premotor cortex and greater extension into the anterior inferior frontal gyrus, especially in the right hemisphere. These results have implications for the evolution of fronto-parietal functions including spatial attention to observed actions, social learning, and tool use, and are in line with previous research suggesting a unique role for the right anterior inferior frontal gyrus in the evolution of human fronto-parietal network architecture.

Novel genotype-phenotype associations in human cancers enabled by advanced molecular platforms and computational analysis of whole slide images.

Technological advances in computing, imaging, and genomics have created new opportunities for exploring relationships between histology, molecular events, and clinical outcomes using quantitative methods. Slide scanning devices are now capable of rapidly producing massive digital image archives that capture histological details in high resolution. Commensurate advances in computing and image analysis algorithms enable mining of archives to extract descriptions of histology, ranging from basic human annotations to automatic and precisely quantitative morphometric characterization of hundreds of millions of cells. These imaging capabilities represent a new dimension in tissue-based studies, and when combined with genomic and clinical endpoints, can be used to explore biologic characteristics of the tumor microenvironment and to discover new morphologic biomarkers of genetic alterations and patient outcomes. In this paper, we review developments in quantitative imaging technology and illustrate how image features can be integrated with clinical and genomic data to investigate fundamental problems in cancer. Using motivating examples from the study of glioblastomas (GBMs), we demonstrate how public data from The Cancer Genome Atlas (TCGA) can serve as an open platform to conduct in silico tissue-based studies that integrate existing data resources. We show how these approaches can be used to explore the relation of the tumor microenvironment to genomic alterations and gene expression patterns and to define nuclear morphometric features that are predictive of genetic alterations and clinical outcomes. Challenges, limitations, and emerging opportunities in the area of quantitative imaging and integrative analyses are also discussed.

Glioblastoma: imaging genomic mapping reveals sex-specific oncogenic associations of cell death.

PURPOSE: To identify the molecular profiles of cell death as defined by necrosis volumes at magnetic resonance (MR) imaging and uncover sex-specific molecular signatures potentially driving oncogenesis and cell death in glioblastoma (GBM). MATERIALS AND METHODS: This retrospective study was HIPAA compliant and had institutional review board approval, with waiver of the need to obtain informed consent. The molecular profiles for 99 patients (30 female patients, 69 male patients) were identified from the Cancer Genome Atlas, and quantitative MR imaging data were obtained from the Cancer Imaging Archive. Volumes of necrosis at MR imaging were extracted. Differential gene expression profiles were obtained in those patients (including male and female patients separately) with high versus low MR imaging volumes of tumor necrosis. Ingenuity Pathway Analysis was used for messenger RNA-microRNA interaction analysis. A histopathologic data set (n = 368; 144 female patients, 224 male patients) was used to validate the MR imaging findings by assessing the amount of cell death. A connectivity map was used to identify therapeutic agents potentially targeting sex-specific cell death in GBM. RESULTS: Female patients showed significantly lower volumes of necrosis at MR imaging than male patients (6821 vs 11 050 mm(3), P = .03). Female patients, unlike male patients, with high volumes of necrosis at imaging had significantly shorter survival (6.5 vs 14.5 months, P = .01). Transcription factor analysis suggested that cell death in female patients with GBM is associated with MYC, while that in male patients is associated with TP53 activity. Additionally, a group of therapeutic agents that can potentially be tested to target cell death in a sex-specific manner was identified. CONCLUSION: The results of this study suggest that cell death in GBM may be driven by sex-specific molecular pathways.

Somatic mutations associated with MRI-derived volumetric features in glioblastoma.

INTRODUCTION: MR imaging can noninvasively visualize tumor phenotype characteristics at the macroscopic level. Here, we investigated whether somatic mutations are associated with and can be predicted by MRI-derived tumor imaging features of glioblastoma (GBM). METHODS: Seventy-six GBM patients were identified from The Cancer Imaging Archive for whom preoperative T1-contrast (T1C) and T2-FLAIR MR images were available. For each tumor, a set of volumetric imaging features and their ratios were measured, including necrosis, contrast enhancing, and edema volumes. Imaging genomics analysis assessed the association of these features with mutation status of nine genes frequently altered in adult GBM. Finally, area under the curve (AUC) analysis was conducted to evaluate the predictive performance of imaging features for mutational status. RESULTS: Our results demonstrate that MR imaging features are strongly associated with mutation status. For example, TP53-mutated tumors had significantly smaller contrast enhancing and necrosis volumes (p = 0.012 and 0.017, respectively) and RB1-mutated tumors had significantly smaller edema volumes (p = 0.015) compared to wild-type tumors. MRI volumetric features were also found to significantly predict mutational status. For example, AUC analysis results indicated that TP53, RB1, NF1, EGFR, and PDGFRA mutations could each be significantly predicted by at least one imaging feature. CONCLUSION: MRI-derived volumetric features are significantly associated with and predictive of several cancer-relevant, drug-targetable DNA mutations in glioblastoma. These results may shed insight into unique growth characteristics of individual tumors at the macroscopic level resulting from molecular events as well as increase the use of noninvasive imaging in personalized medicine.