BIRN Collaboratory Infrastructure

1
BIRN Collaboratory Infrastructure

• Mark James, Project Manager
• BIRN Coordinating Center

August 15, 2004
2
What BIRN is doing

• Integrating the activities of the most advanced
  biomedical imaging and clinical research centers
  in the US - Serving as a model for programs
  everywhere.
• Developing hardware and software infrastructure
  for managing distributed data.
• Exploring data using intelligent query engines
  that can make inferences upon locating
  interesting data.
• Building bridges across tools and data formats.
• Changing the use pattern for research data from
  the individual laboratory/project to shared use.

3
What BIRN is doing

• Define processes, procedures and establish best
  practices so that the BIRN is reliable, scalable
  and extensible to biomedical research programs -
  able to support the work of thousands of
  researchers.
• Push the envelop of biomedical informatics and
  computer science by causing the development of
  new techniques in databases, information
  retrieval, visualization and computational
  processing.

4
What Have We Been Building?
Building a Shared Biomedical IT Infrastructure to
Hasten the Derivation of New Understanding and
Treatment of Disease through the use of
Distributed Knowledge
June 2004
5
BIRN Rack Fully Configured
6
De-Identification of Subjects
Image Header Subject Juan Perez Patient
ID 911
Image Header Subject anon BIRN ID
9284ka9e23sd
BIRN Virtual Data Grid
Health Insurance Portability Accountability Act
(HIPAA) of 1996
7
Brain Morphometry BIRN

• Morphometry BIRN participants are examining
  neuroanatomical correlates of neuropsychiatric
  illnesses in such disorders as Alzheimers
  disease, depression, mild cognitive impairment
  and the ageing brain.
• Through large-scale analyses of patient
  population data acquired and pooled across sites,
  these scientists are investigating whether brain
  structural differences correlate to symptoms such
  as memory dysfunction or depression and whether
  specific structural differences distinguish
  diagnostic categories.
• Increasing the statistical power for studying
  relatively rare populations
• Harvard (MGH and BWH), Duke, UCLA, UC San Diego,
  UC Irvine, Johns Hopkins, Washington University
  in St. Louis

8
MIRIAD Analysis Example

• Deidentified Data from Duke Retrospective Archive
• Loaded in BIRN Data Grid
• UCLA LONI Pipeline
• Register Probabilistic Anatomy Atlas to Subjects
• Lobar Analysis
• BWH/MIT 3D Slicer
• Image Analysis and Segmentation
• UCSD Supercomputers
• Cluster Processing
• Statistical Analysis
• Detailed Clinical Database

BWH Probabilistic Atlas (one time transfer)
BWH Intensity Normalization and EM Segmentation
UCLA AIR Registration and Lobar Analysis
3
2
UCSD Supercomputing
1
4
Duke Archives
Duke Clinical Analysis
MIRIAD Data Flow 1) Retrospective data upload
from Duke 2) Lobar analysis and Registration of
Atlas to Subjects 3) Anatomical Segmentation 4)
Comparison to Clinical History
9
MIRIAD Project Accomplishments
Improved computational capabilities

• Segmentation Duke BIRN-MIRIAD
• Item (semi-automated) (fully-automated)
• of tissue classes 3 (Fig1) 23 (Fig2)
• Time for 200 brains 400 hours 1 hour
• Time for 200 lobe 250 hours all lobes (Fig3)
  and 27 regional analysis regions included
  above

10
MIRIAD Initial Results--Lobar

• Analysis carried out by normalizing regions by
  total brain volume
• 50 depressed, 50 controls, imaged at baseline and
  2 years
• Parietal lobe smaller in depressed (p lt 0.02)
• In subjects responding to therapy Temporal lobe
  smaller (p lt 0.08) Frontal lobe was not smaller
  (p lt 0.6)
• This is the first study to show brain structural
  change over time in response to treatment in
  unipolar depression

11
SASHA Project
4
JHU Shape Analysis of Segmented Structures
3
MGH Segmentation
5
BWH Visualization
UCSD Supercomputing
1
Goal comparison and quantification of
structures shape and volumetric differences
across patient populations
SRB
Data Donor Sites
De-identification And upload
2
12
SASHA Project Accomplishments
Large Deformation Diffeomorphic Metric Mapping
(LDDMM) using the TeraGrid

• Data 46 hippocampus data sets (2070 comparisons)
• Each LDDMM comparison takes about 3 to 8 hours

Single PC TeraGrid
1 comparison 431 days 60 comparisons simultaneously 7 days
Improved computational capabilities
13
Mouse BIRN

• Studying animal models of disease at different
  anatomical scales to test hypothesis associated
  with human neurological disorders
• Share and analyze multi-scale structural and
  functional data and ultimately to integrate them
  with genomic and gene expression data on the
  mouse brain. Ongoing collaborations in basic
  mouse models of neurological collaborations
  disorders include animal models of relevance to
  schizophrenia, Parkinson's disease, brain cancer,
  substance abuse and multiple sclerosis.
• Duke, UCLA, UC San Diego, Cal Tech

14
Mouse BIRN
Integrating brain data across scales and
disciplines
Spatial Registration of Brain Volumes
Reconstructed Spiny Dendrite
UCSD-NCMIR
UCSD-NCMIR Duke - CIVN
UCLA - LONI
15
Mouse BIRN Data Federation
16
Function BIRN

• Developing a common fMRI protocol to study
  regional brain dysfunction related to the
  progression and treatment of schizophrenia
• Calibrating inter-site imaging differences
  between scanner manufacturers
• Correlating functional data with anatomical data
  acquired from the Morphology testbed to study if
  there are neuroanatomical correlates with
  cognitive dysfunction across disorders
• UCLA, UC San Diego, UC Irvine, Harvard (MGH and
  BWH), Stanford, Minnesota, Iowa, New Mexico,
  Duke/U. North Carolina

17
Inter-site variability
How bad it is?
Different scanners different raw images
18
Access to the BIRN Infrastructure
The BIRN shared information technology
infrastructure for basic and translational
research is available to all researchers from any
internet capable location.
19
The BIRN Portal

• Application environment that provides transparent
  and pervasive access to the BIRN infrastructure
  (i.e. tools, applications, resources) with a
  Single Login from any Internet capable location
• Provides simple, intuitive access to distributed
  Grid resources for data storage, distributed
  computation, and visualization
• Provides a scalable interface for users of all
  backgrounds and level of expertise

20
The BIRN Portal

• Provides customized work areas that address
  the common and unique requirements of test bed
  groups and individual users
• Has a flexible architecture built on emerging
  software standards allowing for transparent
  access to sophisticated computational and data
  service

• Requires a minimum amount of administrative
  complexity

21
BIRN Virtual Data Grid

• Defines a Distributed Data Handling System
• Uniform interface for connecting to heterogeneous
  data resources over a network
• Allows for the seamless creation and management
  of distributed data sets

• Virtual file system provides users with a unified
  view of a distributed data collection
• Supports Pervasive Auditing

22
Federated Databases
Are chronic, but not first-onset patients,
associated with superior temporal gyrus
dysfunction (MMN)?
Integrated View
Mediator
Wrapper
Wrapper
Web
Wrapper
Wrapper
Wrapper
Wrapper
PubMed, Expasy
fMRI
Clinical
ERP
Receptor Density
Structure
23
Portal Application Integration
NIH ImageJ FreeSurfer LONI Pipeline 3D
Slicer JViewer SRB Tools Mediator Queries Know-Me
UMLS BIRN Calendar BIRN Message Board LDDMM
JViewer
ImageJ
Slicer
LONI Pipeline
24
BIRN Working Groups
25
Challenges

• Breaking down the barriers
• Mistrust
• Open sharing of information
• Who gets credit
• Commercial products
• Governance
• Incorporating processes for multi-site studies
  and sharing of human data
• HIPAA Compliance
• Patient confidentiality
• Institutional Review Board (IRB) approvals
• Developing guidelines - for sharing data
  authorship
• Integrating new participants
• Providing an architecture to allow for technology
  improvements with the existing infrastructure
• Guaranteeing security versus ease of use

26
Major Accomplishments

• Cultural change
• A new way to do science (biomedical and computer)
  that is both creative and exciting, and works
• Willingness to have others analyze your data with
  their methods
• To facilitate use of your methods on their data
• Sharing concepts as well as data
• Solving common problems rapid turn around time,
  new perspectives

27
http//www.nbirn.net
28
Questions ?