Image mapping, registration and atlases

1
Image mapping, registration and atlases

• Derek Hill
• Imaging Sciences
• School of Medicine
• Kings College London

2
Overview

• Definitions
• Applications
• Medical Research
• Diagnosis
• Therapy planning and guidance
• Drug discovery
• E-science issues
• Breakout group sub-headings

3
Definition of registration

• Determining the transformation,or mapping T that
  relates positions in image A, to positions in a
  second image (or physical space) B.
• When registered, a position x in image A and
  position T(x) in image B are the same position in
  the object.

4
Correspondence

• Registration is a technique for aligning images
  so that corresponding features can be related.
• For image-to-physical space registration, we
  determine correspondence between an image and
  physical positions identified with a 3D localiser

5
Atlases

• Can mean several things
• Single reference subject used to assist in
  analysis of others
• Combination of multiple reference subjects
• intensity average
• Representation of variability
• Pre-labelled dataset used for image segmentation
• Registration required to
• Create atlas if it is from multiple subjects
• Map atlas to patients or research subjects

6
Registration examples

• Multimodality
• Image guided surgery/therapy
• Detecting change over time
• Identifying differences between groups

7
PET-CT registration
MRC cyclotron unit
8
MR-CT registration
CT
MR
CT bone overlaid on MR After affine transformation
9
Multi-modal volume rendering (Ruff 1994)
Hill et al Radiology 191447-454 1994
10
Registration for image guided surgery
11
MRI tumour surface overlaid in microscope
Edwards et al IEEE TMI 191082-1093 2000
12
2D 3D registration for therapy guidance

• MRI x-ray

13
Combining MRI and x-ray

• Case 2. Electrophysiology study and RF ablation.
• 3D multiphase SSFP MR sequence
• 3 phases, 256x256x128, 1.13x1.13x1.0mm3,
  TR3.1ms, TE1.6ms, ?45?
• Tracked biplane x-ray

LAO
AP
14
Registered MRI and catheters
15
Non-rigid registration

• The previous examples have all assumed that the
  mapping has the degrees of freedom of a rigid
  body
• Tissue deformation, image distortion and
  intersubject variability mean more degrees of
  freedom are needed to establish corresondence

16
Pre-contrast
Post-contrast
Subtraction
Post-contrast (rigid registration)
Post-contrast (affine registration)
Post-contrast (non-rigid registration)
Subtraction (rigid registration)
Subtraction (affine registration)
Subtraction (non-rigid registration)
17
MIP rendering
Non-rigid registration
Rigid registration
No registration
Rueckert et al IEEE TMI 18 712-721 1999
18
Intersubject comparisons
8 subject average
Rigid registration
Affine registration
Rueckert, from Medical Image Registration
Hajnal, Hill, Hawkes (eds) CRC Press 2001
19
Intersubject comparison
8 subject average, non-rigid registration using
10mm grid
Rueckert, from Medical Image Registration
Hajnal, Hill, Hawkes (eds) CRC Press 2001
20
Using deformation fields in neuroscience research
Nature, 9 March 2000
21
6 month interval, baseline 2 years prior to
symptoms
Fox et al Lancet 358201-205 2001
22
29 month interval, symptoms appearing
Fox et al Lancet 358201-205 2001
23
5 year interval, symptomatic for 2 years
Fox et al Lancet 358201-205 2001
24
Intersubject comparison by Voxel Based
Morphometry(provided by Colin Studolme, UCSF)
PD MRI
Tissue Segmentation (from PDT2T1)
Regional Tissue Label Density Filter
Regional Gray matter Density
25
Group Comparison of Local Gray Matter Density
Age Matched Normal Group
Test Group FTD or AD
Estimate Warp to Map Each Individual Anatomy to
Common Coordinates
Warp Tissue Density Maps to Common Coordinates
Compare Tissue Density In Common Coordinates
26
E-science issues

• Algorithms run slowly excellent candidates for
  grid services
• Aggregation of data needed to answer medical
  research and drug discovery questions
• Variety of ancillary metadata formats
• Rich and large intrinsic metadata.
• Collaborative working desirable for healthcare
  and research
• Curation currently poor

27
Possible Breakout group sub-headings

• How do we make image registration grid services
  intraoperable?
• Do we need to devise an abstract model for these
  services?
• How should we represent mappings?
• Do we need an ontology?
• Should we use grid-services for a major cross
  validation of algorithms?
• How can, or should,atlases be shared?
• How could these services be used commercially
  (eg for drug discovery)

28
(No Transcript)
29
Results with model
30
Method Optical Tracking

• Registration by optical tracking
• X-ray table c-arm are tracked by Optotrak
• Sliding patient table is tracked by MR system

31
Method Registration Matrix Calculation

• Overall registration transform is composed of a
  series of stages
• Calibration tracking during intervention

32
Other data to register vectors and tensors
http//spl.bwh.harvard.edu8000/pages/ppl/westin/p
apers/smr97/node4.html
Kilner et al Nature 404759-761 2000
33
Cerebral atrophy a macroscopic concomitant of
neurodegeneration

• Alzheimers disease plaques and tangles,
  dendritic, neuronal, synaptic loss... and atrophy
• Advanced disease widespread severe atrophy
• Early disease overlap with normal aging

34
FLUID REGISTRATION
Non-linear, high-dimensional voxel-by-voxel
registration.
Viscous fluid model preserves topology
Regional volume atrophy can be quantified from
the match.
RIGID
FLUID