Modern medical images of the living body astonish, inspire, motivate, and educate the vast majority of biomedical scientists and students working today. With little doubt, digital images of anatomical form and function obtained using advanced imaging devices are some of the most beautiful in all of biology. Because of this digital nature, multi-center biomedical imaging projects or from a collection of smaller efforts, large-scale data archives represent important opportunities for novel research, data mining, and science training. Indeed, archives of neuroimaging data are now acknowledged to represent primary examples of “Big Data” science. Yet, despite considerable efforts to populate them with primary content and meta-data, many existing medical imaging archives fail to consider and provide a compelling user experience. A user is often left to sift through a text-based set of files, links, or search results before ever visualizing a single image.

But now imagine having a compelling, graphical means by which to render, display, and dynamically interact with the content from thousands of biomedical imaging datasets at once. Envision being able to “see” the entirety of an image archive rendered on your computer’s screen as a collection of 3D objects derived directly from the archive’s data. Consider being able to point your mouse at the scene and easily translate, rotate, and re-arrange objects to highlight global patterns, see natural sub-groupings, and identify outliers. Fly-through data, save particularly interesting views, and share your interactions with colleagues as videos or on social media. Click on any object and view the meta-data associated with the imaging data underlying that object. Easily hunt for specific meta-data attributes and have search results shown to you graphically and in real-time. Select, label, color, and save groups of objects by simply selecting them with your fingertip on touchscreen enabled tablet devices. Interactively plot the data to explore basic relationships and differences. Then, with another mouse click, submit the underlying data from your interactions directly to sophisticated data processing workflows for further processing and analysis. All explored visually, using only one’s index finger.

Through this project, we will develop a cutting-edge system for the graphical visualization of complete biomedical data archives. This will be based upon the principles that 1) the imaging data itself can form the basis for data exploration, that 2) visualization of how datasets relate to one another carries essential information, and 3) that well-designed software tools which focus on the user-experience can greatly accelerate the science of big data discovery. Our proposal does NOT concern the creation of or support for a universal database, ontology, or a universally accepted protocol. We will succeed in this highly unique big data visualization R01 project by pursuing the following specific aims:

Specific Aim 1.
Design and Develop a Big Data Visualization Environment with a Particular Focus on the User-Experience: User interactions with Big Data can be efficiently streamlined through the use of interactive means for visualizing data as collections of 3D graphical objects. We will design and construct a web-enabled OpenGL graphical environment which will a) display multiple rendered biomedical imaging-derived objects via a compelling 3D graphical interface; b) put particular emphasis on the user experience, aesthetic attention to detail, permitting dynamic navigation and interaction with up to thousands of biomedical imaging-derived objects simultaneously using only their computer mouse or fingertip.

Specific Aim 2.
Develop an Extensive Library of Graphically-Accessible Means for Visual Mining of Big Data: The visual identification of clusters, sub-groups, outliers, etc is expedited through graphically-driven searches and exploratory data mining. We will facilitate the systematic processing, decomposition, and rendering of thousands of biomedical images by computing and storing with each data set a variety of informative descriptive statistics. We will develop a host of visually attractive graphical displays for use in rapid, visually-driven, data mining, clustering, search, and selection.

Specific Aim 3.
Enable Direct Interactivity with Sophisticated Data Processing Workflow Tools: Manipulation of medical imaging data “by-proxy” as 3D objects will encourage and facilitate subsequent large-scale workflow-based processing, analysis, and modeling. Users will be able to graphically gather 3D objects of interest into groups and/or collections using only their computer mouse or fingertip, then submit the underlying archived data as inputs to a standing or personally constructed library of useful data processing workflows. These workflows will take advantage of remote cluster and cloud-based computing, thereby enabling the streamlined, large-scale, population-level analyses of Big Data.

The NIH has pointedly recognized Big Data visualization as a targeted area of high need. Our project seeks to specifically address the challenge of visualizing large-scale biomedical imaging data through a unique, compelling, and user-focused software solution. Our team is ideally suited to carrying out the specific aims of this proposal and communicating results through peer-reviewed publication and attendance at domestic and international scientific meetings of particular importance for big data. All in all, our proposal seeks to substantially accelerate the ability to turn Big Data into new “discovery science”.

© 2016 Invizian. All Rights Reserved.

For more details and to stay updated on the program development, follow us the NITRC Website also check us out on BitBucket (account required).

The software is available to download for your Windows (invizian_1.6.6_win64_setup) or Linux (invizian_1.4.6-1_amd64.deb) computer, free of charge. Send a note to us here at INVIZIAN with your email address to get access to an example set of data.