SimpleITK Image-Analysis Notebooks: a Collaborative Environment for Education and Reproducible Research.
Yaniv Z, Lowekamp BC, Johnson HJ, Beare R. · 2018
Read SimpleITK Image-Analysis Notebooks: a Collaborative Environment for Education and Reproducible Research. by Yaniv Z, Lowekamp BC, Johnson HJ, Beare R. (2018) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc5959828.txt.
Introduction
Yaniv and co-authors describe SimpleITK image-analysis notebooks as a shared space for teaching and reproducible research. Their paper appeared in the Journal of Digital Imaging. Modern science needs teams to build complex computational workflows. The notebooks rest on the Jupyter web application. They support literate programming, so steps stay documented and repeatable. SimpleITK gives a simpler face to a major image toolkit.
The setup suits scientists with varied programming skill. Online repositories make data sharing easy for users. A testing framework helps keep code sound over time. Many worked examples are free to download from GitHub. They cover segmentation, registration, and other core tasks.
Catalog record
| CREATOR | Yaniv Z, Lowekamp BC, Johnson HJ, Beare R. |
|---|---|
| YEAR | 2018 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 51 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
SimpleITK Image-Analysis Notebooks: a Collaborative Environment for Education and Reproducible Research By Ziv Yaniv, Bradley C Lowekamp, Hans J Johnson, Richard Beare Published in: Journal of Digital Imaging DOI: 10.1007/s10278-017-0037-8 ABSTRACT Abstract Modern scientific endeavors increasingly require team collaborations to construct and interpret complex computational workflows. This work describes an image-analysis environment that supports the use of computational tools that facilitate reproducible research and support scientists with varying levels of software development skills. The Jupyter notebook web application is the basis of an environment that enables flexible, well-documented, and reproducible workflows via literate programming. Image-analysis software development is made accessible to scientists with varying levels of programming experience via the use of the SimpleITK toolkit, a simplified interface to the Insight Segmentation and Registration Toolkit. Additional features of the development environment include user friendly data sharing using online data repositories and a testing framework that facilitates code maintenance. SimpleITK provides a large number of examples illustrating educational and research-oriented image analysis workflows for free download from GitHub under an Apache 2.0 license: github.com/InsightSoftwareConsortium/SimpleITK-Notebooks. Keywords: Image analysis, Open-source software, Registration, Segmentation, R, Python INTRODUCTION Research and educational activities that include analysis of biomedical images are carried out by diverse communities that include physicians, biologists, biomedical engineers, physicists, mathematicians, computer scientists, and others. As a consequence, similar image analysis tasks are performed by people with significantly different levels of software development skills. Less-experienced software developers often use programs that facilitate high-level computational workflows via graphical user interfaces such as 3D Slicer [8] and Fiji [24]. More experienced software developers often prefer to directly use computational toolkits such as the Insight Segmentation and Registration Toolkit (ITK) [13]. Many in this diverse community of scientists have limited formal training and skills as programmers or software developers. This observation was described in [10]: “In the Future, Everyone Will Be a Programmer for 15 Minutes”. To support their work, these non-expert developers need tools to address the challenges of testing and maintaining their software creations. An essential requirement for confirming any conclusions described in a scientific publication, including biomedical image analysis findings, is reproducibility [7, 21]. Satisfying this requirement is challenging when the computational workflow is manual, which is often the case when using programs with graphical user interfaces. In these situations, the workflow is not documented in an automated manner. The need for documented computational workflows is likely the reason why both 3D Slicer and Fiji offer automated scripting capabilities. While this addresses the reproducibility issue, in many cases, the flexibility of such solutions is limited when compared to directly using an image analysis toolkit such as ITK. Conversely, the use of ITK requires considerable proficiency in the C++ object-oriented programming language, a relatively specialized skill. Another significant aspect of modern scientific research is its collaborative nature. Collaborations between multiple groups at different institutions and possibly countries is common and appears to be a hallmark of the current scientific landscape [1, 2]. The work presented here is one such example. Developing biomedical image analysis workflows in a distributed collaborative setting requires sharing of data, computational models in natural language including their mathematical derivations, source code, and the current results both numeric and in graphical form. For active collaborations, all four components need to be synchronized so that participants in the research endeavor are aware of the current state of affairs. In most cases, the four components are separate: (1) there is a document describing the model and mathematical derivation, (2) there are source code files, ideally, managed using a version control system, (3) with both originating data, and (4) summary result files. The management of the separate components is not an ideal approach to communication between participants because it requires that each participant develop connections between sections in the model description, the implementation, and the corresponding results. These issues were clearly identified in the 1980s, with the proposed solution being literate programming [16]. In the literate programming approach, the computation description using natural language, figures, and accompanying mathematical derivations are embedded with the implementation in a single document. This approach facilitates improved understanding in the educational setting and can enhance collaborative work. Unfortunately, while the idea is attractive, it was not widely adopted till rather recently. In this work, we present an image-analysis environment which addresses the challenges described above. We enable flexible reproducible computational workflows using a simplified interface to ITK that is accessible to programmers with varying levels of experience. The workflows are implemented using a web-based literate programming environment that provides opportunities for improving collaborations. Finally, all necessary components are combined in a testable, maintainable image analysis environment with transparent data sharing via online repositories. MATERIALS AND METHODS Three elements that comprise the image-analysis environment are the SimpleITK toolkit, the SimpleITK Jupyter notebooks, and the development infrastructure. SimpleITK Overview SimpleITK, originally introduced in [18], is a simplified programming interface to the algorithms and data structures of the Insight Segmentation and Registration Toolkit. SimpleITK has specializations for multiple programming languages including C+ +, Python, R, Java, C#, Lua, Ruby, and TCL. The SimpleITK interfaces address the primary challenge associated with using ITK algorithms by directly allowing scientific domain experts to use the most common forms of these algorithms in any one of the programming languages that they are familiar with. SimpleITK enables users to use the skills they have rather than requiring a significant amount of effort to become proficient in C+ +. Three characteristics of ITK that make it a phenomenal platform for efficiently implementing complex medical image analysis tools also require advanced software development experience. The first characteristic is that the toolkit is distributed in source code form, requiring the developer to compile ITK before they can use it. The second characteristic derives from ITK’s extensive use of the C+ +template mechanism to provide data-type flexibility. Finally, ITK components are designed for use as part of a data analysis pipeline. ITK elements are connected sequentially to construct the pipeline with no computations performed during pipeline construction. This powerful pipeline architecture adds complexity that is a significant barrier to implementation for many ITK newcomers. Each of these power-enabling characteristics requires specialized understanding of advanced programming concepts. SimpleITK shortens the time it takes to access the ITK toolkit algorithms by trading-off potential flexibility and performance for global ease of use. There are binary distributions for a variety of programming languages. It abandons the use of templates in favor of a small subset of a specific pixel, image, and filter types which correspond to the most widely used data-type combinations. And finally, it hides the use of the filter pipeline in favor of the more simple programming approach whereby the user provides input to a filter which immediately returns the output. When introduced in [18], SimpleITK included a large number of components for image filtering and segmentation. Since then, the toolkit has been significantly extended to include over 280 ITK filters, with the primary addition being the incorporation of the ITK version 4 (ITKv4) registration framework [3]. This addition completes the SimpleITK functionality, with comprehensive support for image filtering, segmentation, and registration. We next describe the new registration framework, starting with its two principal data elements, transformations, and images. Transformations The toolkit supports all