SNI-Abstractx - University of Washington

jumentousmanlyInternet και Εφαρμογές Web

21 Οκτ 2013 (πριν από 3 χρόνια και 9 μήνες)

77 εμφανίσεις


Many areas of science face the problem of understanding and interacting with
imaging data whose scale and scope is relentlessly expanding. The challenge is
particularly acute in neuroimaging and cognitive neuroscience where evolving
instrumentation to meas
ure the brain, along with rapid advances in analysis methods,
have outpaced the scientific infrastructure for validating and modeling the

The Scalable Neuroimaging Initiative (SNI) will support multidisciplinary teams at
University of W

Stanford University

to develop a distributed computing
framework for access to and remote processing of quantitative neuroimaging data. The
project will produce a database system and application programming interface (API)
that is designed sp
ecifically for understanding imaging data.

The SNI team will coordinate software development work at these two sites and
with key stakeholders at other institutions through two workshops that will engage
leaders in neuroimaging, computer science, and in
formatics. The team will develop a
system that exposes quantitative neuroimaging data to image

based query. The project will use

functional and structural connectivity
data as test beds that will be shared among multiple
research sites. Users will be able
to search using a web
based image browser and initiate remote analyses based on
source and extensible algorithms.

Several hundred neuroimaging labs acquire data. A successful database can
allow interactions between

these labs, as well as opening up the data from these labs to
thousands of statisticians, engineers, and mathematicians, thus greatly advancing
scientific knowledge.

Key features of the API will be:

Globally unique identifiers

(GUID). Just as Unifor
m Resource Identifiers (URI’s)
made the semantic web possible, image
level GUIDs make distributed images available
for access and processing via a software layer that accesses images at their source (for
example, from XNAT, Horizon Research Share, or Neuro
biological Image Management
System (NIMS). Retrieved images may be cached locally, but the GUIDs will permit
ambiguous reference to the image sources, no matter where they are on the

Operationalization of anatomico
spatial search semantics

in the query API

example is an ontology web service (OWS) that will provide expert
curated mappings
between specific classes in standard ontologies and labels in standard brain spaces,
thereby fostering interoperability among different brain parcell
ation schemes.

It would
respond to a request for a specific element of its ontology in a (queried dataset) sp
space with a voxel mask

Incorporation of source
agnostic image processing workflow in the query API
, using
source tools based on
NIPy, NIPype, and pyXNAT, which will give access to
workhorse academic image processing packages and databases, including
FreeSurfer , SPM and XNAT
. We will develop workflow abstractions for generating data
quality metrics and verifying image de

Virtualization of the server performing the query computation
, allowing for
parallelization in high performance computing environments (e.g. Microsoft Azure,
Amazon Elastic Compute Cluster). (See Howe LOS, UW eScience).

Contributions to me
tadata standards
, including fields for GUIDs, standard semantics
for time
stamping temporally streamed data, and recording of provenance information
and analysis results using emerging standards like XCEDE and AIM.

Demonstration applications

that use th
is API to create a zero
installation (web
based) query interface, and a 3D interactive viewer.

interface tool evaluations.

The core implementation will include methods for
analyzing user behavior; incorporating software that tracks user
rors and response
time enables improving user interfaces over time.