Efficient and reproducible execution from data collection to processing

研发项目3：从数据收集到处理的高效可复制执行（DO）

TR&D Project 3: Efficient and reproducible execution from data collection to processing (DO) SUMMARY: The ReproNim project seeks to transform neuroimaging practice, to make research more efficient and effective in such a way that also makes it reproducible as a result. As more data, metadata, and computing resources become available to the neuroimaging community, tools and frameworks for managing data and processing workflows that ensure consistent control over all of the digital objects of science become increasingly important. Such tools should assist in obtaining valid results while establishing their provenance and minimizing the need for manual curation and intervention; they should not get in the way of doing research. In this Technology Research and Development Project, TR&D 3, we establish new approaches, as well as adopt and contribute back to existing tools, to automate many stages of data collection and analysis, making efficient use of local or remote computing resources that are available to the researchers. In particular, we aim to 1) Automate “Doing (execution of) an experiment” through collection and representation of data, metadata, and provenance across all stages of a neuroimaging acquisition, including all the data types that could be important for quality assurance and proper accounting for possible confounding factors, such as audio/video stimuli, physiological recordings, details of the experimental design. Automated integration of imaging and non-imaging data not only makes research more efficient and labor saving, it also makes collected and shared data more comprehensive, accurate, and reproducible. 2) Make computational resources (GPUs, local High Performance Computing centers, and cloud computing resources) conveniently and efficiently available to researchers to perform execution of needed data transformations (conversion, analysis, etc.). While orchestrating execution we will record detailed provenance information, sufficient for re-execution of any stage of the research process, and make it available to the researcher alongside with the produced results. Efficient use of computational resources and collection of detailed provenance will facilitate experimentation and application of bleeding edge analysis workflows, while reducing necessary technological know-how. 3) Maintain, support, and extend existing ReproNim and related software and data resources that we and our partners have made available openly to the community. This effort will be complemented by training modules and support for different user experience levels and use cases. Ensuring such continuity in availability and robust operation of tools, computing environments, and data resources is essential for any effort aiming to support efficient and reproducible computation. We will carry out this work in collaboration with the other ReproNim technology research and development projects, our collaborative and service projects, and the neuroimaging community at large. This work will automate and conveniently interface complex technologies while facilitating use of established data standards and provenance recording, lowering the technological expertise necessary for neuroimaging scientists to advance knowledge.