Open access image repositories: high-quality data to enable machine learning research.

Originally motivated by the need for research reproducibility and data reuse, large-scale, open access information repositories have become key resources for training and testing of advanced machine learning applications in biomedical and clinical research. To be of value, such repositories must provide large, high-quality data sets, where quality is defined as minimising variance due to data collection protocols and data misrepresentations. Curation is the key to quality. We have constructed a large public access image repository, The Cancer Imaging Archive, dedicated to the promotion of open science to advance the global effort to diagnose and treat cancer. Drawing on this experience and our experience in applying machine learning techniques to the analysis of radiology and pathology image data, we will review the requirements placed on such information repositories by state-of-the-art machine learning applications and how these requirements can be met.

Genetic variation in social environment construction influences the development of aggressive behavior in Drosophila melanogaster.

Individuals are not merely subject to their social environments; they choose and create them, through a process called social environment (or social niche) construction. When genotypes differ in social environment-constructing behaviors, different genotypes are expected to experience different social environments. As social experience often affects behavioral development, quantitative genetics and psychology theories predict that genetic variation in social environment construction should have an important role in determining phenotypic variation; however, this hypothesis has not been tested directly. I identify multiple mechanisms of social environment construction that differ among natural genotypes of Drosophila melanogaster and investigate their consequences for the development of aggressive behavior. Male genotypes differed in the group sizes that they preferred and in their aggressive behavior; both of these behaviors influenced social experience, demonstrating that these behaviors function as social environment-constructing traits. Further, the effects of social experience-as determined in part by social environment construction-carried over to affect focal male aggression at a later time and with a new opponent. These results provide manipulative experimental support for longstanding hypotheses in psychology, that genetic variation in social environment construction has a causal role in behavioral development. More broadly, these results imply that studies of the genetic basis of complex traits should be expanded to include mechanisms by which genetic variation shapes the environments that individuals experience.

Identifying the social context of single- and mixed-species group formation in large African herbivores.

Despite continued interest in mixed-species groups, we still lack a unified understanding of how ecological and social processes work across scales to influence group formation. Recent work has revealed ecological correlates of mixed-species group formation, but the mechanisms by which concomitant social dynamics produce these patterns, if at all, is unknown. Here, we use camera trap data for six mammalian grazer species in Serengeti National Park. Building on previous work, we found that ecological variables, and especially forage quality, influenced the chances of species overlap over small spatio-temporal scales (i.e. on the scales of several metres and hours). Migratory species (gazelle, wildebeest and zebra) were more likely to have heterospecific partners available in sites with higher forage quality, but the opposite was true for resident species (buffalo, hartebeest and topi). These findings illuminate the circumstances under which mixed-species group formation is even possible. Next, we found that greater heterospecific availability was associated with an increased probability of mixed-species group formation in gazelle, hartebeest, wildebeest and zebra, but ecological variables did not further shape these patterns. Overall, our results are consistent with a model whereby ecological and social drivers of group formation are species-specific and operate on different spatio-temporal scales. This article is part of the theme issue 'Mixed-species groups and aggregations: shaping ecological and behavioural patterns and processes'.

Behavioural syndromes in fishes: a review with implications for ecology and fisheries management.

This review examines the contribution of research on fishes to the growing field of behavioural syndromes. Current knowledge of behavioural syndromes in fishes is reviewed with respect to five main axes of animal personality: (1) shyness-boldness, (2) exploration-avoidance, (3) activity, (4) aggressiveness and (5) sociability. Compared with other taxa, research on fishes has played a leading role in describing the shy-bold personality axis and has made innovative contributions to the study of the sociability dimension by incorporating social network theory. Fishes are virtually the only major taxon in which behavioural correlations have been compared between populations. This research has guided the field in examining how variation in selection regime may shape personality. Recent research on fishes has also made important strides in understanding genetic and neuroendocrine bases for behavioural syndromes using approaches involving artificial selection, genetic mapping, candidate gene and functional genomics. This work has illustrated consistent individual variation in highly complex neuroendocrine and gene expression pathways. In contrast, relatively little work on fishes has examined the ontogenetic stability of behavioural syndromes or their fitness consequences. Finally, adopting a behavioural syndrome framework in fisheries management issues including artificial propagation, habitat restoration and invasive species, may promote restoration success. Few studies, however, have examined the ecological relevance of behavioural syndromes in the field. Knowledge of how behavioural syndromes play out in the wild will be crucial to incorporating such a framework into management practices.

Computer-aided evaluation of neuroblastoma on whole-slide histology images: Classifying grade of neuroblastic differentiation.

Neuroblastoma (NB) is one of the most frequently occurring cancerous tumors in children. The current grading evaluations for patients with this disease require pathologists to identify certain morphological characteristics with microscopic examinations of tumor tissues. Thanks to the advent of modern digital scanners, it is now feasible to scan cross-section tissue specimens and acquire whole-slide digital images. As a result, computerized analysis of these images can generate key quantifiable parameters and assist pathologists with grading evaluations. In this study, image analysis techniques are applied to histological images of haematoxylin and eosin (H&E) stained slides for identifying image regions associated with different pathological components. Texture features derived from segmented components of tissues are extracted and processed by an automated classifier group trained with sample images with different grades of neuroblastic differentiation in a multi-resolution framework. The trained classification system is tested on 33 whole-slide tumor images. The resulting whole-slide classification accuracy produced by the computerized system is 87.88%. Therefore, the developed system is a promising tool to facilitate grading whole-slide images of NB biopsies with high throughput.

Computer-aided Prognosis of Neuroblastoma on Whole-slide Images: Classification of Stromal Development.

We are developing a computer-aided prognosis system for neuroblastoma (NB), a cancer of the nervous system and one of the most malignant tumors affecting children. Histopathological examination is an important stage for further treatment planning in routine clinical diagnosis of NB. According to the International Neuroblastoma Pathology Classification (the Shimada system), NB patients are classified into favorable and unfavorable histology based on the tissue morphology. In this study, we propose an image analysis system that operates on digitized H&E stained whole-slide NB tissue samples and classifies each slide as either stroma-rich or stroma-poor based on the degree of Schwannian stromal development. Our statistical framework performs the classification based on texture features extracted using co-occurrence statistics and local binary patterns. Due to the high resolution of digitized whole-slide images, we propose a multi-resolution approach that mimics the evaluation of a pathologist such that the image analysis starts from the lowest resolution and switches to higher resolutions when necessary. We employ an offine feature selection step, which determines the most discriminative features at each resolution level during the training step. A modified k-nearest neighbor classifier is used to determine the confidence level of the classification to make the decision at a particular resolution level. The proposed approach was independently tested on 43 whole-slide samples and provided an overall classification accuracy of 88.4%.

An Assessment of Imaging Informatics for Precision Medicine in Cancer.

Objectives: Precision medicine requires the measurement, quantification, and cataloging of medical characteristics to identify the most effective medical intervention. However, the amount of available data exceeds our current capacity to extract meaningful information. We examine the informatics needs to achieve precision medicine from the perspective of quantitative imaging and oncology. Methods: The National Cancer Institute (NCI) organized several workshops on the topic of medical imaging and precision medicine. The observations and recommendations are summarized herein. Results: Recommendations include: use of standards in data collection and clinical correlates to promote interoperability; data sharing and validation of imaging tools; clinician's feedback in all phases of research and development; use of open-source architecture to encourage reproducibility and reusability; use of challenges which simulate real-world situations to incentivize innovation; partnership with industry to facilitate commercialization; and education in academic communities regarding the challenges involved with translation of technology from the research domain to clinical utility and the benefits of doing so. Conclusions: This article provides a survey of the role and priorities for imaging informatics to help advance quantitative imaging in the era of precision medicine. While these recommendations were drawn from oncology, they are relevant and applicable to other clinical domains where imaging aids precision medicine.

Coordinated cell motility is regulated by a combination of LKB1 farnesylation and kinase activity.

Cell motility requires the precise coordination of cell polarization, lamellipodia formation, adhesion, and force generation. LKB1 is a multi-functional serine/threonine kinase that associates with actin at the cellular leading edge of motile cells and suppresses FAK. We sought to understand how LKB1 coordinates these multiple events by systematically dissecting LKB1 protein domain function in combination with live cell imaging and computational approaches. We show that LKB1-actin colocalization is dependent upon LKB1 farnesylation leading to RhoA-ROCK-mediated stress fiber formation, but membrane dynamics is reliant on LKB1 kinase activity. We propose that LKB1 kinase activity controls membrane dynamics through FAK since loss of LKB1 kinase activity results in morphologically defective nascent adhesion sites. In contrast, defective farnesylation mislocalizes nascent adhesion sites, suggesting that LKB1 farnesylation serves as a targeting mechanism for properly localizing adhesion sites during cell motility. Together, we propose a model where coordination of LKB1 farnesylation and kinase activity serve as a multi-step mechanism to coordinate cell motility during migration.

Image-guided genomics of phenotypically heterogeneous populations reveals vascular signalling during symbiotic collective cancer invasion.

Phenotypic heterogeneity is widely observed in cancer cell populations. Here, to probe this heterogeneity, we developed an image-guided genomics technique termed spatiotemporal genomic and cellular analysis (SaGA) that allows for precise selection and amplification of living and rare cells. SaGA was used on collectively invading 3D cancer cell packs to create purified leader and follower cell lines. The leader cell cultures are phenotypically stable and highly invasive in contrast to follower cultures, which show phenotypic plasticity over time and minimally invade in a sheet-like pattern. Genomic and molecular interrogation reveals an atypical VEGF-based vasculogenesis signalling that facilitates recruitment of follower cells but not for leader cell motility itself, which instead utilizes focal adhesion kinase-fibronectin signalling. While leader cells provide an escape mechanism for followers, follower cells in turn provide leaders with increased growth and survival. These data support a symbiotic model of collective invasion where phenotypically distinct cell types cooperate to promote their escape.

Classification of small cell lung cancer and pulmonary carcinoid by gene expression profiles.

Small cell lung cancer is a common type of lung cancer that is generally classified within the spectrum of neuroendocrine lung neoplasms. Using high-density cDNA arrays, we profiled gene expression of small cell lung cancers and compared these expression profiles to those of normal bronchial epithelial cells and pulmonary carcinoids, which are classified as benign neuroendocrine tumors. We found the overall expression profiles of two small cell lung cancer cell lines, two microdissected tissue samples of primary small cell lung cancer, and cultured bronchial epithelial cells to be relatively similar to one another, with an average Pearson correlation coefficient for these comparisons of 0.63. However, we found the expression profiles of small cell lung cancers (and bronchial epithelial cells) to be surprisingly dissimilar to those of two samples of pulmonary carcinoid tumors, with an average correlation coefficient for these comparisons of 0.20. We then compared the pulmonary carcinoid expression profiles to those of two samples of infiltrating astrocytic brain cancers (oligodendroglioma and high-grade astrocytoma) and found similarity of gene expression among these four samples (average correlation coefficient, 0.57). These gene expression profiles suggest that small cell lung cancers are closely related to (and possibly derived from) epithelial cells, and that pulmonary carcinoids are related to neural crest-derived brain tumors. More generally, our results suggest that broad profiles of gene expression may reveal similarities and differences between tumors that are not apparent by traditional morphological criteria.

The prognostic significance of recall antigen testing in melanoma patients.

A quantitative assessment of the long-term prognostic value and clinical usefulness of recall antigen reactions in patients with malignant melanoma is not available. The authors evaluated longitudinal observations of survival made in 846 patients over a 12-year period. Each patient was initially studied with Mantoux-type recall antigen skin tests. The patients were categorized with respect to the following: high (greater than 5 mm) or low (less than or equal to 5 mm) averaged skin test reaction diameters at 48 hr; Clark level; tumor stage (I = localized tumor, II = local extension and/or region lymph node metastasis, III = systemic metastasis); ulceration; site of primary; histologic type; age; and sex. The percentage of high reactors in Stages I, II, and III were 44.3%, 37.4%, and 25%, respectively. Survival was evaluated with the Cox-Mantell hazard function model and the Cox regression model. The significant (chi-squared; probability) risk factors detected were tumor stage (94.58; less than or equal to 0.0001), Clark level (19.37; less than or equal to 0.0001), sex (16.97; less than or equal to 0.0001), and skin test reactivity (7.48; less than or equal to 0.0062). A significant relationship also was detected between skin test reactor status and the tumor stage (p less than or equal to 0.0330). When evaluated within each stage of disease, skin test reactivity predicted survival only in Stage II patients (p less than or equal to 0.0080). Five-year survival estimates among Stage II patients were 58% among high reactors and 38% among low reactors.(ABSTRACT TRUNCATED AT 250 WORDS)

A proprotein convertase/MMP-14 proteolytic cascade releases a novel 40 kDa vasculostatin from tumor suppressor BAI1.

Brain-specific angiogenesis inhibitor 1 (BAI1), an orphan G protein-coupled receptor-type seven transmembrane protein, was recently found mutated or silenced in multiple human cancers and can interfere with tumor growth when overexpressed. Yet, little is known about its regulation and the molecular mechanisms through which this novel tumor suppressor exerts its anti-cancer effects. Here, we demonstrate that the N terminus of BAI1 is cleaved extracellularly to generate a truncated receptor and a 40-kDa fragment (Vasculostatin-40) that inhibits angiogenesis. We demonstrate that this novel proteolytic processing event depends on a two-step cascade of protease activation: proprotein convertases, primarily furin, activate latent matrix metalloproteinase-14, which then directly cleaves BAI1 to release the bioactive fragment. These findings significantly augment our knowledge of BAI1 by showing a novel post-translational mechanism regulating BAI1 activity through cancer-associated proteases, have important implications for BAI1 function and regulation, and present novel opportunities for therapy of cancer and other vascular diseases.

Supercomputers and biological sequence comparison algorithms.

Comparison of biological (DNA or protein) sequences provides insight into molecular structure, function, and homology and is increasingly important as the available databases become larger and more numerous. One method of increasing the speed of the calculations is to perform them in parallel. We present the results of initial investigations using two dynamic programming algorithms on the Intel iPSC hypercube and the Connection Machine as well as an inexpensive, heuristically-based algorithm on the Encore Multimax.

The Virtual Microscope.

We present the design of the Virtual Microscope, a software system employing a client/server architecture to provide a realistic emulation of a high power light microscope. We discuss several technical challenges related to providing the performance necessary to achieve rapid response time, mainly in dealing with the enormous amounts of data (tens to hundreds of gigabytes per slide) that must be retrieved from secondary storage and processed. To effectively implement the data server, the system design relies on the computational power and high I/O throughput available from an appropriately configured parallel computer.

Development of a universal connectivity and data management system.

Point-of-care testing (POCT) is an increasingly popular method of delivering laboratory testing. Management of POCT is challenging given the variety of devices, locations, and staff that need to be coordinated to ensure quality results and meet regulatory guidelines. Electronic capture and transfer of data are preferred for managing POCT, but there is currently no standard method of connecting different devices. Johns Hopkins Medical Institutions (JHMI) developed a common data management system with interfaces to all of its POCT devices. All POCT data are collected in one database and analyzed in a similar fashion. Where data were once collected by carrying laptops to each nursing unit, the POCT devices can now connect directly to the database over the Internet. Algorithms have been created to automate the data analysis and review process. Over the several years that this software has been used, JHMI has experienced improved quality, accuracy, and management of its POCT program. The labor saved by increased automation of data review is refocused on enhancing the performance and scope of the program. Current connectivity and analysis algorithms have future application to remote consultation, management of home self-monitoring patients, and examination of real-time data.

A graphical tool for ad hoc query generation.

Medical data are characterized by complex taxonomies and evolving terminology. Questions that clinicians, medical administrators, and researchers may wish to answer using medical databases are not easily formulated as SQL queries. In this paper we describe a graphical tool that facilitates formulation of ad hoc questions as SQL queries. This tool manages multiple attribute hierarchies and creates SQL query strings by navigating through the hierarchies. This interactive tool has been optimized using indexing to improve the overall speed of the query building and the data retrieval process. Indexed queries performed 5 to 100 times faster than query strings. However, query string generation time depends on the size of the taxonomies describing the hierarchies, while the index generation time depends on the size of the data warehouse.

Digital dynamic telepathology--the Virtual Microscope.

The Virtual Microscope is being designed as an integrated computer hardware and software system that generates a highly realistic digital simulation of analog, mechanical light microscopy. We present our work over the past year in meeting the challenges in building such a system. The enhancements we made are discussed, as well as the planned future improvements. Performance results are provided showing the system scales well, so that many users can be adequately serviced by an appropriately configured data server.