National Cancer Institute Imaging Data Commons: Toward Transparency, Reproducibility, and Scalability in Imaging Artificial Intelligence.

The remarkable advances of artificial intelligence (AI) technology are revolutionizing established approaches to the acquisition, interpretation, and analysis of biomedical imaging data. Development, validation, and continuous refinement of AI tools requires easy access to large high-quality annotated datasets, which are both representative and diverse. The National Cancer Institute (NCI) Imaging Data Commons (IDC) hosts large and diverse publicly available cancer image data collections. By harmonizing all data based on industry standards and colocalizing it with analysis and exploration resources, the IDC aims to facilitate the development, validation, and clinical translation of AI tools and address the well-documented challenges of establishing reproducible and transparent AI processing pipelines. Balanced use of established commercial products with open-source solutions, interconnected by standard interfaces, provides value and performance, while preserving sufficient agility to address the evolving needs of the research community. Emphasis on the development of tools, use cases to demonstrate the utility of uniform data representation, and cloud-based analysis aim to ease adoption and help define best practices. Integration with other data in the broader NCI Cancer Research Data Commons infrastructure opens opportunities for multiomics studies incorporating imaging data to further empower the research community to accelerate breakthroughs in cancer detection, diagnosis, and treatment. Published under a CC BY 4.0 license.

NCI Imaging Data Commons.

The National Cancer Institute (NCI) Cancer Research Data Commons (CRDC) aims to establish a national cloud-based data science infrastructure. Imaging Data Commons (IDC) is a new component of CRDC supported by the Cancer Moonshot. The goal of IDC is to enable a broad spectrum of cancer researchers, with and without imaging expertise, to easily access and explore the value of deidentified imaging data and to support integrated analyses with nonimaging data. We achieve this goal by colocating versatile imaging collections with cloud-based computing resources and data exploration, visualization, and analysis tools. The IDC pilot was released in October 2020 and is being continuously populated with radiology and histopathology collections. IDC provides access to curated imaging collections, accompanied by documentation, a user forum, and a growing number of analysis use cases that aim to demonstrate the value of a data commons framework applied to cancer imaging research. SIGNIFICANCE: This study introduces NCI Imaging Data Commons, a new repository of the NCI Cancer Research Data Commons, which will support cancer imaging research on the cloud.

Generalized Linear Models outperform commonly used canonical analysis in estimating spatial structure of presence/absence data.

BACKGROUND: Ecological communities tend to be spatially structured due to environmental gradients and/or spatially contagious processes such as growth, dispersion and species interactions. Data transformation followed by usage of algorithms such as Redundancy Analysis (RDA) is a fairly common approach in studies searching for spatial structure in ecological communities, despite recent suggestions advocating the use of Generalized Linear Models (GLMs). Here, we compared the performance of GLMs and RDA in describing spatial structure in ecological community composition data. We simulated realistic presence/absence data typical of many ß-diversity studies. For model selection we used standard methods commonly used in most studies involving RDA and GLMs. METHODS: We simulated communities with known spatial structure, based on three real spatial community presence/absence datasets (one terrestrial, one marine and one freshwater). We used spatial eigenvectors as explanatory variables. We varied the number of non-zero coefficients of the spatial variables, and the spatial scales with which these coefficients were associated and then compared the performance of GLMs and RDA frameworks to correctly retrieve the spatial patterns contained in the simulated communities. We used two different methods for model selection, Forward Selection (FW) for RDA and the Akaike Information Criterion (AIC) for GLMs. The performance of each method was assessed by scoring overall accuracy as the proportion of variables whose inclusion/exclusion status was correct, and by distinguishing which kind of error was observed for each method. We also assessed whether errors in variable selection could affect the interpretation of spatial structure. RESULTS: Overall GLM with AIC-based model selection (GLM/AIC) performed better than RDA/FW in selecting spatial explanatory variables, although under some simulations the methods performed similarly. In general, RDA/FW performed unpredictably, often retaining too many explanatory variables and selecting variables associated with incorrect spatial scales. The spatial scale of the pattern had a negligible effect on GLM/AIC performance but consistently affected RDA's error rates under almost all scenarios. CONCLUSION: We encourage the use of GLM/AIC for studies searching for spatial drivers of species presence/absence patterns, since this framework outperformed RDA/FW in situations most likely to be found in natural communities. It is likely that such recommendations might extend to other types of explanatory variables.

Open Health Imaging Foundation Viewer: An Extensible Open-Source Framework for Building Web-Based Imaging Applications to Support Cancer Research.

PURPOSE: Zero-footprint Web architecture enables imaging applications to be deployed on premise or in the cloud without requiring installation of custom software on the user's computer. Benefits include decreased costs and information technology support requirements, as well as improved accessibility across sites. The Open Health Imaging Foundation (OHIF) Viewer is an extensible platform developed to leverage these benefits and address the demand for open-source Web-based imaging applications. The platform can be modified to support site-specific workflows and accommodate evolving research requirements. MATERIALS AND METHODS: The OHIF Viewer provides basic image review functionality (eg, image manipulation and measurement) as well as advanced visualization (eg, multiplanar reformatting). It is written as a client-only, single-page Web application that can easily be embedded into third-party applications or hosted as a standalone Web site. The platform provides extension points for software developers to include custom tools and adapt the system for their workflows. It is standards compliant and relies on DICOMweb for data exchange and OpenID Connect for authentication, but it can be configured to use any data source or authentication flow. Additionally, the user interface components are provided in a standalone component library so that developers can create custom extensions. RESULTS: The OHIF Viewer and its underlying components have been widely adopted and integrated into multiple clinical research platforms (e,g Precision Imaging Metrics, XNAT, LabCAS, ISB-CGC) and commercial applications (eg, Osirix). It has also been used to build custom imaging applications (eg, ProstateCancer.ai, Crowds Cure Cancer [presented as a case study]). CONCLUSION: The OHIF Viewer provides a flexible framework for building applications to support imaging research. Its adoption could reduce redundancies in software development for National Cancer Institute-funded projects, including Informatics Technology for Cancer Research and the Quantitative Imaging Network.

LesionTracker: Extensible Open-Source Zero-Footprint Web Viewer for Cancer Imaging Research and Clinical Trials.

Oncology clinical trials have become increasingly dependent upon image-based surrogate endpoints for determining patient eligibility and treatment efficacy. As therapeutics have evolved and multiplied in number, the tumor metrics criteria used to characterize therapeutic response have become progressively more varied and complex. The growing intricacies of image-based response evaluation, together with rising expectations for rapid and consistent results reporting, make it difficult for site radiologists to adequately address local and multicenter imaging demands. These challenges demonstrate the need for advanced cancer imaging informatics tools that can help ensure protocol-compliant image evaluation while simultaneously promoting reviewer efficiency. LesionTracker is a quantitative imaging package optimized for oncology clinical trial workflows. The goal of the project is to create an open source zero-footprint viewer for image analysis that is designed to be extensible as well as capable of being integrated into third-party systems for advanced imaging tools and clinical trials informatics platforms. Cancer Res; 77(21); e119-22. ©2017 AACR.

Applying the dark diversity concept to nature conservation.

Linking diversity to biological processes is central for developing informed and effective conservation decisions. Unfortunately, observable patterns provide only a proportion of the information necessary for fully understanding the mechanisms and processes acting on a particular population or community. We suggest conservation managers use the often overlooked information relative to species absences and pay particular attention to dark diversity (i.e., a set of species that are absent from a site but that could disperse to and establish there, in other words, the absent portion of a habitat-specific species pool). Together with existing ecological metrics, concepts, and conservation tools, dark diversity can be used to complement and further develop conservation prioritization and management decisions through an understanding of biodiversity relativized by its potential (i.e., its species pool). Furthermore, through a detailed understanding of the population, community, and functional dark diversity, the restoration potential of degraded habitats can be more rigorously assessed and so to the likelihood of successful species invasions. We suggest the application of the dark diversity concept is currently an underappreciated source of information that is valuable for conservation applications ranging from macroscale conservation prioritization to more locally scaled restoration ecology and the management of invasive species.

Long-term impacts of nitrogen deposition on coastal plant communities.

Nitrogen deposition has been shown to have significant impacts on a range of vegetation types resulting in eutrophication and species compositional change. Data from a re-survey of 89 coastal sites in Scotland, UK, c. 34 years after the initial survey were examined to assess the degree of change in species composition that could be accounted for by nitrogen deposition. There was an overall increase in the Ellenberg Indicator Value for nitrogen (EIV-N) of 0.15 between the surveys, with a clear shift to species characteristic of more eutrophic situations. This was most evident for Acid grassland, Fixed dune, Heath, Slack and Tall grass mire communities and despite falls in EIV-N for Improved grass, Strand and Wet grassland. The increase in EIV-N was highly correlated to the cumulative deposition between the surveys, and for sites in south-east Scotland, eutrophication impacts appear severe. Unlike other studies, there appears to have been no decline in species richness associated with nitrogen deposition, though losses of species were observed on sites with the very highest levels of SOx deposition. It appears that dune vegetation (specifically Fixed dune) shows evidence of eutrophication above 4.1 kg N ha(-1) yr(-1), or 5.92 kg N ha(-1) yr(-1) if the lower 95% confidence interval is used. Coastal vegetation appears highly sensitive to nitrogen deposition, and it is suggested that major changes could have occurred prior to the first survey in 1976.

Species composition of coastal dune vegetation in Scotland has proved resistant to climate change over a third of a century.

Climate change is expected to have an impact on plant communities as increased temperatures are expected to drive individual species' distributions polewards. The results of a revisitation study after c. 34 years of 89 coastal sites in Scotland, UK, were examined to assess the degree of shifts in species composition that could be accounted for by climate change. There was little evidence for either species retreat northwards or for plots to become more dominated by species with a more southern distribution. At a few sites where significant change occurred, the changes were accounted for by the invasion, or in one instance the removal, of woody species. Also, the vegetation types that showed the most sensitivity to change were all early successional types and changes were primarily the result of succession rather than climate-driven changes. Dune vegetation appears resistant to climate change impacts on the vegetation, either as the vegetation is inherently resistant to change, management prevents increased dominance of more southerly species or because of dispersal limitation to geographically isolated sites.

Development of a bent Laue beam-expanding double-crystal monochromator for biomedical X-ray imaging.

The Biomedical Imaging and Therapy (BMIT) beamline at the Canadian Light Source has produced some excellent biological imaging data. However, the disadvantage of a small vertical beam limits its usability in some applications. Micro-computed tomography (micro-CT) imaging requires multiple scans to produce a full projection, and certain dynamic imaging experiments are not possible. A larger vertical beam is desirable. It was cost-prohibitive to build a longer beamline that would have produced a large vertical beam. Instead, it was proposed to develop a beam expander that would create a beam appearing to originate at a source much farther away. This was accomplished using a bent Laue double-crystal monochromator in a non-dispersive divergent geometry. The design and implementation of this beam expander is presented along with results from the micro-CT and dynamic imaging tests conducted with this beam. Flux (photons per unit area per unit time) has been measured and found to be comparable with the existing flat Bragg double-crystal monochromator in use at BMIT. This increase in overall photon count is due to the enhanced bandwidth of the bent Laue configuration. Whilst the expanded beam quality is suitable for dynamic imaging and micro-CT, further work is required to improve its phase and coherence properties.

Altered lung motion is a sensitive indicator of regional lung disease.

Since lung diseases adversely affect airflow during breathing, they must also alter normal lung motion, which can be exploited to detect these diseases. However, standard imaging techniques such as CT and MRI imaging during breath-holds provide little or no information on lung motion and cannot detect diseases that cause subtle changes in lung structure. Phase-contrast X-ray imaging provides images of high contrast and spatial resolution with temporal resolutions that allow multiple images to be acquired throughout the respiratory cycle. Using X-ray phase-contrast imaging, coupled with velocimetry, we have measured lung tissue movement and determined velocity fields that define speed and direction of regional lung motion throughout a breath in normal Balb/c nude male mice and mice exposed to bleomycin. Regional maps of lung tissue motion reveal both the heterogeneity of normal lung motion, as well as abnormal motion induced by bleomycin treatment. Analysed histologically, bleomycin treatment caused pathological changes in lung structure that were heterogenous, occupying less than 12% of the lung at 6 days after treatment. Moreover, plethysmography failed to detect significant changes in compliance at either 36 h or 6 days after treatment. Detailed analysis of the vector fields demonstrated major differences (p < 0.001) in regional lung motion between control and bleomycin-treated mice at both 36 h and 6 days after treatment. The results of this study demonstrate that X-ray phase-contrast imaging, coupled with velocimetry, can detect early stage, subtle and non-uniform lung disease.

Simultaneous acquisition of dual analyser-based phase contrast X-ray images for small animal imaging.

Analyser-based phase contrast X-ray imaging can provide high-contrast images of biological tissues with exquisite sensitivity to the boundaries between tissues. The phase and absorption information can be extracted by processing multiple images acquired at different analyser orientations. Recording both the transmitted and diffracted beams from a thin Laue analyser crystal can make phase retrieval possible for dynamic systems by allowing full field imaging. This technique was used to image the thorax of a mechanically ventilated newborn rabbit pup using a 25 keV beam from the SPring-8 synchrotron radiation facility. The diffracted image was produced from the (111) planes of a 50 mm x 40 mm, 100 microm thick Si analyser crystal in the Laue geometry. The beam and analyser were large enough to image the entire chest, making it possible to observe changes in anatomy with high contrast and spatial resolution.

Fourier transform infrared imaging and small angle x-ray scattering as a combined biomolecular approach to diagnosis of breast cancer.

Fourier transform infrared (FTIR) microspectroscopic imaging and small angle x-ray scattering (SAXS) were combined to investigate the supramolecular structure of collagen from 27 tissue sections from patients undergoing mastectomy, excisional biopsy, or mammoplasty. Both techniques were correlated by matching the scattering profile from the SAXS data with the integrated area of the infrared collagen region (1300-1180 cm(-1)). The FTIR spectral profiles and multivariate analysis of various tissue components showed consistent differences between all major tissue components, particularly between cancer and normal tissue cells. Analysis of the SAXS data revealed broad differences between cancer and normal tissue, but were inconclusive due to the small sample size. Parameters were extracted from each technique in relation to their characterization of collagen to reveal a good correlation between the two techniques, which diagnostically parallels with gold-standard Hematoxylin and Eosin (H&E) stained sections. The results show that the integrated area of collagen region in the FTIR spectrum for cancerous samples is greater than that for noncancerous samples indicating collagen disorder. This supports the notion that collagen is structurally disrupted in cancer tissue consistent with the interpretation of the SAXS data. Overall, both these techniques successfully distinguished cancer from normal breast tissue. Integration of these two techniques was able to better segregate cancer as well as provide a more complete understanding of the differences in collagen on all structural levels during breast cancer development.

Microscale vapour diffusion for protein crystallization.

The development of new crystallization platforms via the application of high-throughput technologies has delivered a plethora of crystallization plates suitable for robot-driven and manual setups. However, practically all these plates (except for microfluidic channel chips) are based on a very similar design and well (precipitant):drop (protein) volume ratios. A new type of crystallization plate (microplate) has therefore been developed and tested that still employs the classical vapour-diffusion technique but minimizes the precipitant well volume to 1.2 microl for a 150 nl protein drop setup. This enables a very significant saving on the total bulk of the crystallization screen, hence allowing the application of new, rare and expensive solutions in automated crystallization-screening procedures. Additionally, owing to the very low drop:well volume ratio, the new microplate can significantly accelerate the equilibrium time necessary for crystal nucleation and growth, in many cases shortening the high-throughput crystallization screening process to a few hours.

Analyser-based mammography using single-image reconstruction.

We implement an algorithm that is able to decode a single analyser-based x-ray phase-contrast image of a sample, converting it into an equivalent conventional absorption-contrast radiograph. The algorithm assumes the projection approximation for x-ray propagation in a single-material object embedded in a substrate of approximately uniform thickness. Unlike the phase-contrast images, which have both directional bias and a bias towards edges present in the sample, the reconstructed images are directly interpretable in terms of the projected absorption coefficient of the sample. The technique was applied to a Leeds TOR[MAM] phantom, which is designed to test mammogram quality by the inclusion of simulated microcalcifications, filaments and circular discs. This phantom was imaged at varying doses using three modalities: analyser-based synchrotron phase-contrast images converted to equivalent absorption radiographs using our algorithm, slot-scanned synchrotron imaging and imaging using a conventional mammography unit. Features in the resulting images were then assigned a quality score by volunteers. The single-image reconstruction method achieved higher scores at equivalent and lower doses than the conventional mammography images, but no improvement of visualization of the simulated microcalcifications, and some degradation in image quality at reduced doses for filament features.