Atlas Registration Services and Workflows within the INCF Digital Atlasing Infrastructure for Rodent Brain

Ilya Zaslavsky (San Diego Supercomputer Center, University of California San Diego, USA), Steven Lamont (University of California San Diego, USA), Asif Memon (University of California San Diego, USA), Luis Ibanez (Kitware, Inc., USA), James Gee (University of Pennsylvania, USA), Stephen Larson (University of California San Diego, USA), David Little (University of California San Diego, USA), David Valentine (San Diego Supercomputer Center, University of California San Diego, USA), Richard Baldock (MRC Human Genetics Unit and IGMM, UK), Mike Hawrylycz (Allen Institute for Brain Science, USA), Janis Breeze (INCF, Sweden), Raphael Ritz (INCF, Sweden), Jyl Boline (Informed Minds, USA)

The International Neuroinformatics Coordinating Facility (INCF) has been developing standards-based infrastructure supporting interoperability of atlas data (INCF Digital Atlasing Infrastructure, or DAI). As part of this work, Waxholm Markup Language (WaxML), an output schema for exchanging location-specific atlas information, and INCF Atlas Services, a collection of web service functions for retrieving atlas data based on location and translating location information between formally defined atlas coordinate systems, have been specified. Several INCF atlas hubs, hosting the atlas data services compliant with the specification, have been developed. The next step is enabling easy inclusion of images, reconstructions and other types of atlas data in INCF-DAI. Its key component is developing web services for registering images to one or more canonical spatial frameworks defined in INCF-DAI, and organizing the services in workflows that address a range of registration use cases.

Several research groups have implemented software tools for 2D and 3D spatial registration such as: a large collection of procedures is available in the ITK/VTK software suite; landmark-based spatial registration is implemented in the Mouse Brain Atlasing Toolkit (MBAT); a workflow for warping and registering large images (multi-Gb) using a compute cluster has been implemented at University of California San Diego (UCSD); spatial registration procedures are implemented in the Edinburgh Map Atlas Project (EMAP). However, these efforts are fragmented and often tightly coupled with their specific client environments. Further, they rely on different representation of data and spatial transformations which precludes their joint use. We report on an effort to systematize spatial registration procedures, develop common service interfaces for them, allow them to be integrated into flexible registration workflows, and make them accessible from multiple clients within INCF-DAI in a manner compatible with the atlas services and WaxML.

The registration services prototype is built from several components: ITK/VTK/ANTS scripts wrapped as web services compatible with the OGC Web Processing Service (WPS) standard (the same standard that atlas services follow); Jibber and Jetsam software tools for registration of large images also wrapped in WPS service interfaces; a set of standard formats for exchanging 2D and 3D images and transformation descriptions that are shared between the ITK suite and the Jibber/Jetsam registration workflow; a set of interfaces with the underlying common data grid environment (based on iRODS), which supports seamless storage management and storage extension when required. The standard services support: (1) segmenting initial images and representing them in a common packaged format for registration, (2) registering images of selected modalities to the Waxholm spatial reference system by establishing landmark pairs, (3) generating a standard description of the spatial transformation based on the landmarks, or using another method available in the ITK/VTK suite, (4) using the derived transformation description to generate a warped image (using high performance compute resources if required), and (5) making the registered image and its associated spatial reference and transformation information available via atlas services API. As a result of the registration workflow, the spatial reference system (SRS) of the registered image is added to the SRS registry maintained by INCF-DAI, while the transformation between WHS and the new image SRS is added to the INCF-DAI registry of transformations, thus enabling spatial integration of data in the new image with atlas data available elsewhere in the system.

 

Preferred presentation format: Demo
Topic: Digital atlasing

Document Actions