Transcript if Reliable & Easy To Use

XCEDE Use Cases:
QueryAtlas & other scenarios
2008 fBIRN AHM
Overview
•
User Centered Design for fBIRN tools
•
Example Users of Slicer with informatics infrastructure:
•
fBIRN researchers (QueryAtlas & population statistics)
•
Neuroscience labs (image analysis workflow)
•
Retrospective data sharing (mark-up)
•
CVRG (statistics & multi-scale information integration)
•
Image-guided therapy (real-time image access)
•
Scenarios
•
Requirements
User-centered design approach
Drive tool development with user needs, use cases –
ensure that XCEDE web services address use cases of
interest.
1. Assess user + developer needs,
priorities & develop model
2, 3, 4: Design, Implement, Test, Iterate:
* Users evaluate and shape the design (features +
interaction) that works for them.
* Developers ensure the design’s technical merit, assess
community interest & compatibility with priorities.
5: Disseminate & support
Use case: fBIRN Neuroscientists
•
Using a webGUI or a Slicer query mechanism:
Search for all SZ subjects that have FreeSurfer
+ FIPS analyses for a particular experiment.
•
Or find SZ population statistical analyses and
FreeSurfer average morphology data
•
View downloaded results in QueryAtlas’s
ontology- and atlas-based context; choose
search terms and query for relevant multi-scale
information.
•
View integrated search results & save all data of
interest for continued study.
•
Export descriptions/data compatible with other
atlases
•
Query must be simple & powerful. Tools must
manage & display returned results for clarity.
Use case: Neuroscience labs
•
Example: Schizophrenia Research
•
Structural Imaging, DTI, fMRI
•
Demographics, Protocols, Clinical
Assessments
•
•
Label Maps, Activation Maps
Labs often use Ad Hoc methods for data transfer,
description & tracking provenance
•
Shared Network Drive, Access Database,
Spreadsheets…
•
Reluctant to Change if Database Administrator is
Required
•
Would Greatly Benefit from Informatics
Infrastructure if Reliable & Easy To Use
Use case: Sharing Retrospective Data
•
Over 1000 Slicer-Guided Open MR Craniotomies
•
Large Subset (~200) Cases Collated with Preand Intra-Operative Images
•
•
Various acq., 1.5T and 0.5T
•
GE or NRRD format (not DICOM)
National Center for Image Guided Therapy
(Jolesz, BWH) Goal to Share the Data
•
Talos, Archip, and Haker Working with Marcus et
al. for Upload to XNAT Central
•
Mark-up of image data and derived data
•
Prostate Brachytherapy Data Also Available for
Upload
Use case: CVRG
•
Cardiovascular Research Grid (Winslow, JHU)
•
Broad Range of Potential inputs: CT, MR,
EKG, Pet, DWI…
•
Initial Three Cases of CT Torso Cases
Uploaded to XNAT Central
•
Data from Drs. Triedman and Jolley,
Boston Childrens Pediatric Cardiology
•
Includes Detailed Heart Segmentation
•
JHU Canine Heart DTI
(Want to use QueryAtlas-style ontology- and
atlas-based query for related multi-scale data.)
Use case: Image Guided Therapy
•
Collect Pre-Therapy Imaging
•
Run Diverse Analysis Routines
•
•
Evolving Research Code
•
Multiple Investigators
Bring Patient Maps into Real Time
Navigation / Robotic System
•
Feed Procedure Images and Data Back to
Database
Scenario 1: Manual Per-Subject
Target Clinical Study Scenario:
1. Search XNAT or HID for Target Data (via web GUI or Slicer queries)
- Search for project name or subject name of interest:
- Search for particular data assigned to you for that project
2. Select datasets of interest for download.
3. Download .xcat file containing URIs for selected datasets
4. Slicer3 Retrieves Data and Metadata
5. Analysis, Visualization, Annotation
6. Select particular datasets for upload and tag them
7. New Data and Provenance is Uploaded and Linked to Original Data
Scenario 1: Manual Per-Subject
XCEDE Working Group discussion:
Some return fields could also be included:
find_subjects request returns a list of subjects and
This: find_subjects(project="OASIS", fields="CDR,Age,Dx") Returns
their IDs.
get_subjects could be used to retrieve details in
this:
<List>
<Item id="1234" type="subject">
XCEDE format.
<Field name="CDR">0</Field>
<Field name="Age">87</Field>
<Field name="Dx">Normal</Field>
This: find_subjects(project=“PNL-PROJECT")
</Item>
Returns this:
<Item id="3456" type="subject">
<List>
<Field name="CDR">0.5</Field>
<Item id="1234" type="subject"/>
<Field name="Age">84</Field>
<Item id="3456" type="subject"/>
<Field name="Dx">AD</Field>
<Item id="5678" type="subject"/>
</Item>
</List>
<Item id="5678" type="subject">
<Field name="CDR">0</Field>
<Field name="Age">65</Field>
<Field name="Dx">Normal</Field>
</Item>
</List>
Scenario 1: Manual Per-Subject
Slicer MRML representation for holding results in scene state:
vtkMRMLSlicerTableNode &
vtkMRMLSlicerTableStorageNode (M. Halle)
Parse and store data in table form…
Subject ID
Age
CDR
Dx
1234
87
0
Normal
3456
84
0.5
5678
65
0
AD
Normal
•
Also allows useful CSV input format for collections of URIs
•
Any URI table data handed off to appropriate vtkMRMLURIHandler
Scenario 2: Population Analysis
Target Population Study Scenario:
1. Search XNAT or HID Target Population (How will people want to
specify search?)
2. Receive .xcat File of Result URIs
3. Search or Load Parameter Files for Tools – intuitive UI or script
4. Specify Group Statistics Parameters – intuitive UI or script
5. Dispatch Analysis for Remote Processing – intuitive UI or script
6. Results are Uploaded to Database from Remote Analysis Machines
 Identify fBIRN tool users who can help us flesh out this scenario &
understand user needs?
Scenario 3: Visualization of Statistics
Target Visualization Scenario:
1. Search XNAT or HID Processed Studies for diagnosis (SZ),
protocol (SIRP), that have both FreeSurfer and FIPS analyses.
2. Select among returned results for studies of interest.
3. Receive .xcat File of Result URIs
4. Parse & load into Slicer / QueryAtlas (tags distinguish data
types/formats for intelligent loading)
5. Query for explanatory information
6. Add notes, save query record and relevant information as part
of MRML scene
7. Upload MRML scene to database
Requirements
•
Easy To Install XCEDE Compliant Server
•
•
XNAT Desktop?
Web Services API and Web services client tools
•
The Simpler the Better
•
Several Examples are Available
•
•
DigitalBucket, GeneNetwork
Capability to represent data provenance, and mark up data with
custom tags (Assigned To, Completed By)…
•
Good user feedback (is transfer running? Was it successful?)
•
Server-to-Server Data Migration
•
Automatic Synchronization?
•
cvs/svn-like checkouts and commits?