Warum Brauchen Wir Medical Image Computing?

1
Warum Brauchen Wir Medical Image Computing?

• Ron Kikinis, M.D.
• Professor of Radiology,
• Harvard Medical School

Founding Director, Surgical Planning Laboratory,
Brigham and Womens Hospital Principal
Investigator, National Alliance for Medical Image
Computing (a National Center for Biomedical
Computing, part of the Roadmap Initiative), and
Neuroimage Analysis Center (a NCRR National
Resource Center) Research Director, Image Guided
Therapy Program, Brigham and Womens Hospital
2
Acknowledgments

• F. Jolesz, C. Tempany, A. Golby, S. Wells, N.
  Hata, CF. Westin, M. Halle, S. Pieper, F. Talos,
  W. Schroeder, and many more.

3
Medical Imaging

• Medical Imaging is a very powerful tool to look
  into the body
• Non-invasive
• Of increasing importance
• Used both in diagnostics and treatment
• Increasingly powerful and complex

4
MIC The Problem

• More image data, more complexity
• Medical Image Computing aims to extract relevant
  information from images

5
The power of image analysis
A. Golby, P. Black, R. Kikinis
6
Example A Clinical Case

• Right handed male patient, 20 years old.
• Scan of the head after sport trauma

7
Background on Neuroimaging
8
MR images Brain Morphology
Cross-sectional images through the head
A. Golby, P. Black, R. Kikinis
9
MR Images T1 and T2 Weighted
T1 weighted images fat is bright T2 weighted
images water is bright
A. Golby, P. Black, R. Kikinis
10
Diffusion Weighted Imaging (DWI)
http//www.scielo.br/scielo.php?scriptsci_arttext
pidS0004-282X2005000200011
Use MRI to measure Diffusion in tissues
Blue
Red
http//pubs.niaaa.nih.gov/publications/arh27-2/IMA
GES/Page148.gif
11
Diffusion Weighted Imaging (DWI)
Diffusion is represented as a tensor DWI measures
components of this tensor The diffusion tensor is
estimated from those components
S. Pujol, R. Gollub
12
Diffusion Tensors
A. Golby, P. Black, R. Kikinis
13
Connecting Tensors to Form Tracts
A. Golby, P. Black, R. Kikinis
14
Tractography
A. Golby, P. Black, R. Kikinis
15
MR images Tractography
A. Golby, P. Black, R. Kikinis
16
Background on Neurofunction
17
Cortico Spinal Tract

• The cortico-spinal tract connects the primary
  motor cortex to the body. The corticospinal
  pathway is indispensable for moving the fingers
  when reaching and manipulating.

From http//nawrot.psych.ndsu.nodak.edu/Courses/Ps
ych465.S.02/Movement/Brain.html
R. Kikinis
18
Language Network simple version

• Wernickes area language comprehension with
  input from auditory or visual cortex areas
  (depending on the source).
• Brocas area stores motor programs for speaking
  words (and other functions)
• The arcuate fasciculus is a white matter tract
  connecting Wernicke's and Broca's areas

From http//www.sciencecases.org/mini_aphasia/bac
kground.asp
R. Kikinis
19
A Clinical Case Overview
20
A Clinical Case White Matter
21
A Clinical Case Peritumoral Tracts
22
A Clinical Case Virtual Probing
23
A Clinical Case Virtual Probing
24
Research in the OR

• Co-existence of research and clinical
  environments
• Need for interoperation

25
System Integration
Slicer3
BrainLab
BioImage Suite
OpenIGTLink
VVLink
A. Golby, N. Hata, H. Liu
26
Slicer3
A. Golby, N. Hata, H. Liu
27
Intraoperative Navigation
A. Golby, N. Hata, H. Liu
28
Summary

• Use of imaging and image post processing for
  surgical planning and intra-operative navigation
• Difficult location close to critical structures
• Good post-operative results No neurologic
  deficits and no recurrence until now

29
MIC The Science

• Algorithm research (computer science)
• Software tool development
• Biomedical research (applications)?

Courtesy R. Jose et al.
Courtesy P. Black et al.
30
The Software

• Software delivers algorithmic technology to
  biomedical researchers.
• Bridges the gap between algorithms and biomedical
  applications
• Delivery vehicle for advanced clinical care

31
MIC software

• Research software empowers researchers by
• Reducing duplication
• Lowering barriers for scientific exchange
• Clinical software needs to be more
• Robust,
• Better documented and
• Easier to use

32
Research Software

• Free Open Source Software is a good approach for
  research software
• Who will pay for it?
• The taxpayer! Companies will also contribute if
  the software is of value for them (e.g. Linux)
• We have adopted a BSD style open source license
  model

33
The NA-MIC Kit A Modular Platform

• Applications
• Toolkits
• Software Process and Framework

CTestCDash / CPack
CMake
BatchMake
34
NA-MIC Kit Numbers

• Outreach and Community Support
• Public web servers 10 million web hits/month
• Approximate Downloads
• Slicer 5,000 / year
• VTK 7,000 / month
• ITK 5,000 / month
• CMake 700 / day (gt200,000/year)
• Numbers do not include CVS/SVN access and are
  not curated.
• Community size
• Measured in tens of thousands of users
• Hundreds of active developers (250)

35
3D Slicer
36
Contributing Centers

• NA-MIC Develops a free open source software
  platform for analyzing biomedical images
• NAC Leverage the platform for Neuroimage
  analysis
• NCIGT Adds specific IGT capabilities
• BIRN Contributes image informatics
• Harvard Catalyst Usability

37
Community Building

• Training
• National and International Events
• MIT, MGH, UNC, EPFL, NIH, UNM, UCSD, TUM
• All Materials on Wiki
• Clinical and Technical Tracks
• Workshops
• MICCAI 2005, 2006, 2007, 2008..
• OHBM, RSNA, Munich, NCI

38
Community Building

• Working weeks
• Eight So Far (June 2005 January 2009)
• Summers MIT Stata Center
• Winters Salt Lake City
• gt 100 Participants
• Universities Companies
• Wiki and Telephone Preparation
• Project Teams Sit and Work Together
• Face to Face Communication

39
Clinical Software

• Regulatory requirements are put in place to
  protect patients
• Compliance is expensive
• A pure open source model does not work for
  commercial clinical settings
• Open source platforms as a basis for proprietary
  solutions

40
Conclusions

• Imaging is a non-invasive window into the body
• Imaging devices produce more images and more
  complex images
• MIC is critical for extracting information from
  medical imaging data

41
Challenges Today

• Many algorithms are not sufficiently robust
• Translational research needs product-like
  software that is open
• This does not fit into current funding mechanisms

42
Fraunhofer MEVIS

• Goals
• Covers
• image based diagnosis and procedures
• from basic over applied research to products
• Embedded in worldwide clinical network
• Interested in solutions for particular medical
  problems which have high epidemiological impact