Application Programming Interface (API) for Connectome Analysis through Joint Annotation of Large 3D Data (CAJAL3D)
Recent advancements in high-throughput electron microscopy (EM) allow for imaging hundreds of cubic microns of cortical tissue at nanometer-scale resolution. The resulting data afford neuroscientists the opportunity to reverse-engineer cortical microcircuits at an unprecedented level of detail. However, the scale of these data necessitates considerable systems engineering and infrastructure to store, analyze, and share data products efficiently. For example, a cortical column at the resolution of currently available large datasets (e.g., 4 x 4 x 45-nm voxels) is on the order of 1 petabyte when stored on disk. To address this issue, CAJAL3D API has been developed by JHU/APL to implement the Reusable Annotation Markup for Open coNnectomes (RAMON) data model, expose network data services, and enable distributed processing. The purpose of this software package is to make all of the resources and services provided by the CAJAL3D framework easily available to a researcher or software developer.
Paper: R. Burns et al., "The Open Connectome Project Data Cluster: Scalable Analysis and Vision for High Throughput Neuroscience"
CAJAL3D works with several main components of the Open Connectome Project:
- The RAMON model, which is an object model to standardize data products (specifically annotations) and facilitate the interoperability of algorithms and analysis between different institutions. This model defines a set of data types and associated metadata fields that can be used to describe and annotate massive 3D EM image volumes.
- Web services to expose an underlying optimized 3D spatial database that stores both image and RAMON-compliant annotation data. Examples of the services provided include arbitrary 3D cuboid cutouts of image and annotation data, creation and manipulation of RAMON objects, and spatial and metadata queries of RAMON objects.
- A distributed processing framework built on the LONI Pipeline to facilitate large-scale automated analysis of EM data.
- Visualization of EM image data and RAMON annotations via Rambo3D, a graphics processing unit (GPU)-based 3D viewer, and CATMAID, an open-source web-based 2D viewer.
No-Cost Software License
In an effort to promote and facilitate this research, JHU/APL is granting nonexclusive, no-cost rights to CAJAL3D API for individuals and organizations working in fields supporting the U.S. government, education, and other nonprofit purposes.