Satrajit S. Ghosh, Ph.D. - US grants

[unreadable] DESCRIPTION (provided by applicant): The general aim of this project is to disseminate software that will enhance the quality and consistency of analysis of functional magnetic resonance imaging (fMRI) data. The goal is to enhance, document and make publicly available software for artifact detection, statistical region-of-interest analysis and visualization of fMRI data. Better quality control methods and statistical methods will generate more credible and repeatable results, which should therefore lead to faster biomedical discoveries and to potential reduction in the cost of running fMRI studies. From a software engineering standpoint, the goal is to offer a well-designed, cross- platform, extensible software that is intuitive and easy to use. Two existing MATLAB-based software packages will be enhanced, integrated and distributed: the ARtifact detection Tools (ART) and the Region of Interest Analysis of Parcellated Imaging Data (RAPID) software (Nieto-Castanon et al., 2003). To achieve interoperability, the integrated software will be converted from MATLAB to C/C++ and wrappers will be provided for use of this software from other languages such as Python, Java, Tcl/Tk and MATLAB. Custom modules will be created for use of this software within some functional analysis streams (FMRIB Software Library, FSL, Smith et al., 2004, Statistical Parametric Mapping, SPM, Friston 2003, FreeSurfer Functional Analysis STream, FSFAST, Tsao et al., 2003 and Neuroimaging in Python, NiPy), and support will be provided to embed the software in other analysis streams. To achieve dissemination, the software will be beta-tested at several laboratories doing fMRI research and will be maintained and supported through the Neuroimaging Informatics Tools and Resources Clearinghouse (NITRC) website. The project will be carried out in three phases: (1) Integration and release of MATLAB-based software; (2) Conversion of the software to a C/C++ framework; and (3) Dissemination and support to ensure widespread use by the neuroimaging community. Throughout the project we will engage with the neuroimaging community through the NITRC website and, in particular, interact with the several laboratories that have committed to beta-testing the software. We will rely on community feedback to improve usability of the software. PUBLIC HEALTH RELEVANCE: The proposed project aims to disseminate software for sophisticated statistical analyses (Nieto-Castanon et al., 2003) and quality control of functional magnetic resonance imaging (fMRI) data. Providing these tools should enable users of fMRI technology to produce more detailed, consistent and reliable results. This will lead to better understanding of how the brain works and thereby directly impact approaches to diagnosing and treating neurological disorders. [unreadable] [unreadable]

Project Summary With the tremendous increase of neuroimaging data, there is a corresponding demand for usable, automated, and robust data analysis tools. Nipype is a mature Python library for efficient and flexible analysis of Big ?brain imaging? Data. Its reusable workflows can combine algorithms from a diverse set of existing software packages to generate reproducible results. The goal of this proposal is to further enhance the usability, functionality, and interoperability of Nipype and to widen its dissemination. This will increase its use by researchers and clinicians, boost its impact on biomedical research, and address many of its current limitations. Easier-to-use automation tools can reduce errors, lead to faster biomedical discoveries, and facilitate the transition from bench to bedside. From a software engineering standpoint, the goal is to offer a well-designed, cross-platform, and extensible dataflow computing solution that is intuitive and easy to use. We propose to build an interactive and intuitive web-based platform on top of the current extensive feature set of Nipype that interoperates with existing databases, software, and other workflow services. The result will be a generalizable, scalable, extensible, and tested infrastructure that minimizes complex programming interfaces to easier-to-use web applications. Nipype will still retain its extensible plugin architecture behind this web- based platform to allow continued inclusion of new software packages and algorithms, and execution on multiple platforms. Users will be able to use the most appropriate analysis strategies for their data. This platform will not only allow continued use of familiar software, but provide immediate exposure to the latest software tools for data analyses. For analysis, users will have access to complete provenance allowing others to reproduce their steps. We will interact with NeuroVault and NeuroSynth to provide a seamless transition between data, processing, sharing, and interpreting results. Finally, to sustain such an open and collaborative effort, we will train users and developers through hands-on workshops and webinars, encouraging them to take advantage of an expanding ecosystem for efficient and reproducible analysis. While the architecture will be initially deployed within the brain imaging community, we will adopt common standards to ensure interoperability with the greater biomedical imaging community. By continuing to engage the user community and extending the ecosystem for research computing, the project will lower the barrier for easy and efficient computation on large datasets, with the goal of faster development of treatment options.

Neuroscientific data contain information from an incredible diversity of species, are generated by a plethora of devices, and encapsulate the results of scientific thinking and decision making. Most of this generated data remains confined within laboratories and is not accessible to the broader scientific community. The research projects awarded under the Brain Initiative are generating a diverse collection of data that can transform and accelerate the pace of discovery. These datasets are large--ranging in size from GBs to PBs-- and represent diverse data types and assorted metadata. To integrate, rather than further isolate, these numerous efforts there is a need to archive, preserve, share, and process data in a way that is meaningful to neuroscience researchers. Any technological solution should reduce redundancy of storage and computation, allow computing near data, and provide easy, but protected when appropriate, access to researchers or citizen scientists. Given the scale of these initiatives and the range of sample sizes and data types, any solution should also consider the broad range of individual technical expertise in the community and therefore allow easy engagement with and ingestion into an archive, while supporting education and training of the scientists in using these technologies. To solve these problems, we propose ?DANDI: Distributed Archives for Neurophysiology Data Integration.?We leverage our team?s extensive experience in informatics, standards development, software engineering, community building, and leverage a robust open-source software stack to create this archive. The archive will lower barriers for neuroscientists by using the ?Neurodata Without Borders (NWB; ?http://nwb.org?) standard as a consistent data format, by providing interoperability with other standards, and by providing robust tools and convenient Web interfaces to interact with the archive. DANDI will: 1) ?provide a cloud platform for versioned neurophysiology data storage for the purposes of collaboration, archiving, and preservation. 2) ?provide easy to use tools for neurophysiology data submission and access in the archive; and 3) facilitate adoption of NWB via standardized applications for data ingestion, visualization and processing. ?We will work with local investigators, the broader neurophysiology community, and with federal and other funders to determine how long and which pieces of data will be stored in DANDI. The archive will also use state of the art data distribution technologies to increase redundancy and fault tolerance, and allow distributed computing across cloud and local computing resources. Consequently the effort will significantly reduce the barrier between laboratories and the cloud, fostering collaboration and data exchange. Overall, we aim to leverage our collective expertise to create and support an NWB-based neurophysiology archive that seamlessly integrates with and enhances current researcher workflows, lowers barriers for scientific inquiry and collaboration, and preserves information for wide reuse.

There is an increasing need for efficient and robust software to process, integrate, and offer insight across the diversity of population imaging efforts underway across the BRAIN Initiative and other projects. Advances in statistical learning offer a set of technologies that can address many research applications using the extensive and varied data being produced by the projects. This can transform how we analyze and integrate new data. We propose using Nobrainer, an open source Python library that leverages these new learning technologies, as a platform that greatly simplifies integrating deep learning into neuroimaging research. Using this library, we are building and distributing user-friendly and cloud enabled end-user applications for the neuroimaging community. In Aim 1, we provide neural network models. We will create robust, pre-trained neural networks for brain segmentation and time series processing using brain scans from over 65000 individuals. Once trained, these models can then be used as the basis for many other applications, especially in reducing time of processing. We will subsequently use these base networks to perform image processing, image correction, and quality control. In Aim 2, we address the ability to train on private datasets. We will use Bayesian neural network models, which support principled use of prior information. We will use these networks to help detect when the models are expected to fail on an input, and provide visualizations to better understand how the model is working. In Aim 3, we focus on the engineering needed to maintain the software infrastructure, improve efficiency, and increase the scalability of our training methods. Here, we will extend, maintain, and disseminate Nobrainer, our open source software framework, together with training materials and ready to use, cloud-friendly, applications. We will also create much faster, neural network equivalents of time consuming image processing tasks (e.g., registration, segmentation, and annotation). The Nobrainer tools developed through these aims will allow users to find and apply the most pertinent applications and developers to extend the framework to support new architectures and disseminate new models and applications. We expect these tools to be used by any neuroimaging researcher through integration with BRAIN archives and popular software packages. These tools will significantly reduce data processing and new model development time, thus allowing faster exploration of hypotheses using public data and increase reusability of data through greater trust in model outputs.