BISQUE: Enabling Cloud and Grid Powered Image Analysis Ramona Walls iPlant Collaborative Adapted from slides by Martha Narro, Nirav Merchant
Motivation I High throughput imaging is essential for large-scale phenotyping. Affordable robotics for image acquisition are creating vast amounts of imaging data. Many laboratories have automated imaging setups, but lack a comparable analysis platform. Super resolution microscopy and multi-channel images are pushing the boundaries of storage and computational capabilities.
Motivation II New, improved analysis algorithms are being published. Biologists struggle to use them. Developers need images to test algorithms. Scientists need to compare algorithms, reproduce results. Metadata is key for managing large datasets. Sharing and collaborating with large image data sets is challenging.
Bisque Image Management, Analysis, Sharing System
Why Bisque? Biologists can Manage images Choose from multiple analysis options Overlay results to validate findings Annotate images Share images, results, annotations via secure link Algorithm developers can Publish new analysis methods, easily make them web accessible Produce interactive plots, visualizations using built in API Integrated with iPlant storage and computation infrastructure for scalability
How does it work? Bisque iPlant Data Store High Bandwidth Transfer iPlant Computational Infrastructure High Bandwidth Transfer
Bisque Features Web application Tiling, zooming, step through image stacks, play as movie Display 20K x 20K pixel images in web browser Handles 100+ image, video formats Import large image sets (≤ 40 GB via Bisque), extremely large ones (> 40 GB via iPlant Data Store) Scale analyses using distributed computing (connected to XSEDE) and workflow engines (Pegasus, Condor)
Import and Export support for multiple files plugin free HTML5 uploads drag/drop import annotations processing compressed files compose 5D from multiple files on the file streaming export dataset export export annotations compressed files: Tar, GZip, Zip, BZip
Sharing Easy modes: “Private” and “Published” Google style sharing for resources Using s of collaborators Read-only or Full access Easy modes: “Private” and “Published” Google style sharing for resources Using s of collaborators Read-only or Full access
Example Analysis: Seed Size High resolution flat bed scanner image of seeds Edge detection and analysis by Bisque Source: Edgar Spalding
ImageJ modules Bisque API for ImageJ Dataset Cluster/parallel execution Run image macros/plugins using Web-interface Currently requires hand editing of macro files. Working on semi-automated conversion
Bisque-iPlant Team Bisque (U. California, Santa Barbara) B. S. Manjunath Kris Kvelikval Dmitry Fedorov iPlant (U. Arizona, Tucson) Nirav Merchant Martha Narro
Main application: bisque.iplanctc.org BISQUE video tutorials: elp Support Useful Links