Interactive Analysis of Terabyte-sized SEM-EDS Hyperspectral Images Antoine Vandecreme 1 , Peter Bajcsy 1 , Nicholas W.M. Ritchie 2 , and John Henry J. Scott 2 1. Information Technology Lab, National Institute of Standards and Technology, Gaithersburg, MD. 2. Material Measurement Lab, National Institute of Standards and Technology, Gaithersburg, MD. Scanning electron microscopy with energy-dispersive X-ray spectrometry (SEM-EDS) was used to characterize a macroscopic artifact from Peru (a Moche cast figure) [1]. A 3D hyperspectral datacube was acquired over 51.2 hours, measuring 11,520 x 9,984 pixels with 2,048 energy channels per pixel; each pixel corresponds to 900 nm in the x and y dimensions. With current technology, interacting with this hyperspectral dataset to perform exploratory analysis, displaying gigapixel elemental maps, and surveying terabyte-scale spectral data is problematic. We addressed the problem of interacting with terabyte-sized SEM-EDS hyperspectral images by extending the Deep Zoom approach used for large 2D images outside of the field of microanalysis. This approach is based on multi-resolution pyramid representations of 2D images with large pixel counts, and allows for efficient transmission and viewing of images [2]. The initial 2D support was extended to 3D for medical image volumes [3] and to other informative visualizations [4] with an open source project leading to many additional functionalities (see the OpenSeaDragon project [5]). We leveraged the OpenSeaDragon project [5] for building functionality plug-ins. We focused on enabling interactive visual inspections and measurements by supporting real-time interaction with terabyte-sized 3D data, performing off-line re-projections and providing on-demand scale bars and X-ray emission line information overlays. In order to allow further processing and exploration of smaller subsets of terabyte-sized 3D volumes, we added sub-sampling capabilities, executed either client side (i.e in the web browser) or server side, that include a file with provenance information about the sub-sampled data set. To support these functions, we created pyramid representations of the hyperspectral volume that include X-Y and re-projected X-ray spectral views of the dataset, as illustrated in Figure 1 (right). The pyramid set representation corresponding to the X-Y view consists of 4,890,627 input pyramid files in 32,768 folders. The pyramid set for the X-ray spectral view is composed of 4,976,641 output pyramid files in 184,320 folders. Figure 2 shows some of the interactive capabilities accessible in a web browser for the X-ray spectral view. The current capabilities have been deployed in the NIST webspace [6][7][8]. References: [1] DE Newbury and NWM Ritchie, J. Anal. At. Spectrom. 28 (2013), p. 973. [2] Microsoft, “Deep Zoom Silverlight,” Microsoft Developer Network (MSDN), 2010. [Online]. http://msdn.microsoft.com/en-us/library/cc645050(v=vs.95).aspx. [Accessed: 27-May-2013]. [3] S Saalfeld, A Cardona, V Hartenstein, and P Tomančák in “The Collaborative Annotation Toolkit for Massive Amounts of Image Data (CATMAID),” (Max Planck Institute of Molecular Cell Biology and Genetics, Pfotenhauerstr. 108 01307 Dresden, Germany), 2013. [4] “Cosmic chronology ChronoZoom,” University of California, Berkeley, 2013. [Online]. http://www.chronozoomproject.org/BehindTheScenes.htm. [Accessed: 08-Aug-2013]. [5] “Open Seadragon” Open Seadragon project, 2013. [Online].http://openseadragon.codeplex.com/ [6] http://isg.nist.gov/moche/microscopeWebVisualization.html 654 doi:10.1017/S1431927614004991 Microsc. Microanal. 20 (Suppl 3), 2014 © Microscopy Society of America 2014