Neuroradiology Series: Brain Tumors Advanced Imaging Techniques

1
RSNA 2007 Multisession Course VN31
Neuroradiology Series Brain Tumors Advanced
Imaging Techniques
Diffusion and Diffusion Tensor Imaging
Aaron S. Field, MD, PhD Department of
Radiology University of WisconsinMadison
afield_at_uwhealth.org
2
Diffusion and Diffusion Tensor Imaging

• Principles
• Technical Caveats
• Applications and Limitations for Tumor Imaging

3
Diffusion Brownian motion Random molecular
displacements
4
Diffusion probes tissue microstructure
ADC Apparent Diffusion Coefficient
More barriers Fewer barriers ADC low
ADC high
5
Barriers to diffusion in white matter results in
directional dependence (anisotropy)
Anisotropic diffusion
Isotropic diffusion
Fiber tract bundle
6
Anisotropic Signal Attenuation
Diffusion gradient
signal
c/o A. Alexander, UW-Madison
7
The Diffusion Ellipsoid Model
Tensor ellipsoid model estimates the actual ADC
profile
8
The Diffusion Ellipsoid Model
Actual ADC profile (unknown)
Tensor ellipsoid model estimates the actual ADC
profile
9
The Tensor Matrix
Solve matrix equations
xx
yy
zz
xy
xz
xy
10
Rotating the Coordinate Frame
11
Eigenvalues

• Diffusivity parallel to fibers D ?1
• Diffusivity perpendicular to fibers D? (?2
  ?3)/2
• Mean Diffusivity (ADC) (?1 ?2 ?3)/3
• Anisotropy S.D.(?1 , ?2 , ?3)

12
Anisotropy Magnitude of Directionality

• Fractional Anisotropy (FA)
• Most common anisotropy measure
• Summarizes disparity of eigenvalues

13
Tensor Orientation
Courtesy K. Hasan, UTMS-Houston
The major eigenvector (e1) can be mapped using
RGB color coding
14
Tensor Orientation
Brightness FA
Brightness FA Color e1
15
Fiber Tracking (Tractography)
c/o M. Lazar, UW-Madison
Fiber Assignment by Continuous Tracking FACT
(Mori et al)
Although computer graphical trajectories or
streamlines are typically called fibers they
do not correspond to actual nerve fibers!
16
Stopping Criteria for Tractography

• Anisotropy below threshold not reliable
• e.g. stop if FA lt 0.2
• Excessively sharp turn not realistic
• e.g. stop if trajectory takes gt 60 turn

17
Parsing Specific Tracts
1
2
2
1 ROI 2 ROIs
18
Quantitative Fiber Tracking
Tract-specific Regions of Interest
c/o M. Lazar, UW-Madison
19
Quantitative Fiber Tracking
Fiber Trajectory Count
Thomas et al, Brain 2005
20
Quantitative Fiber Tracking
Fiber Density Index of trajectories per voxel
within ROI
(Roberts et al, AJNR 2005)
21
Diffusion and Diffusion Tensor Imaging

• Principles
• Technical Caveats
• Applications and Limitations for Tumor Imaging

22
Anisotropy is Mathematically Non-Specific
23
Anisotropy is Mathematically Non-Specific
Anisotropy ? Tensor Shape!
?FA
Alexander et al, Magn Reson Med 2000
24
Anisotropy is Mathematically Non-Specific
Anisotropy ? Tensor Shape!
Prolate Oblate
Westin et al, ISMRM 1997 Kindlmann et al, Vis
1999 Alexander et al MRM 2000
25
Partial Volume Effects on Anisotropy
26
Partial Volume Effects on Anisotropy
27
Partial Volume Effects on Anisotropy
FA T2
28
Crossing Fibers Limit Tracking to Lateral Motor
Strip
29
Number of trajectories depends on technique
and arbitrary stopping criteria
30
Number of trajectories depends on technique
and arbitrary stopping criteria
FA gt 0.20
31
Number of trajectories depends on technique
and arbitrary stopping criteria
FA gt 0.18
32
False Positives (Spurious Tracts)
33
Fiber count / density
Sensitive to disease both within and distant from
ROI
FDi 2
34
Fiber count / density
Sensitive to disease both within and distant from
ROI
FDi 1
35
Fiber count / density
Sensitive to disease both within and distant from
ROI
FDi 2
36
Fiber count / density
Sensitive to disease both within and distant from
ROI
FDi 1
37
Fiber count / density
FDi 2
38
Fiber count / density
May be decreased despite no loss of tracts within
ROI
FDi 1
39
Fiber count / density
May be normal despite reduced anisotropy within
ROI
FDi 2
40
Small metastasis with extensive vasogenic edema
c/o A. Bizzi, Milan, Italy
41
Diffusion and Diffusion Tensor Imaging

• Principles
• Technical Caveats
• Applications and Limitations for Tumor Imaging

42
Applications of Diffusion and Diffusion
Tensor Imaging to Brain Tumors

• Tissue Characterization (DWI / DTI)
• Discriminate tumor from non-neoplastic lesions
• Estimate tumor histology and grade
• Monitor treatment response
• Discriminate edema from tumor infiltration (?)
• Localization / Mapping (DTI)
• Localize lesions to specific WM tracts
• Map WM tracts for pre-op planning intra-op
  guidance
• Assess integrity / relocation of WM tracts post-op

43
Applications of Diffusion and Diffusion
Tensor Imaging to Brain Tumors

• Tissue Characterization (DWI / DTI)
• Discriminate tumor from non-neoplastic lesions
• Estimate tumor histology and grade
• Monitor treatment response
• Discriminate edema from tumor infiltration (?)
• Localization / Mapping (DTI)
• Localize lesions to specific WM tracts
• Map WM tracts for pre-op planning intra-op
  guidance
• Assess integrity / relocation of WM tracts post-op

44
Epidermoid
Arachnoid cyst
T1C T2 DWI
45
Abscess
GBM
T1C T2 DWI
46
Counterexample GBM with ?ADC
T1C DWI
ADC
47
Counterexample Abscess without ?ADC
T1C T2
DWI
48
ADC inversely correlated with tumor cellularity
Sugahara et al, JMRI 1999
Kono et al, AJNR 2001
49
ADC discriminates pediatric posterior fossa tumors
Rumboldt et al. AJNR 2006
50
ADC may discriminate high/low tumor grade
Sugahara et al, JMRI 1999
Sadeghi et al, AJR 2003
51
Glioblastoma Multiforme
T2 DWI
ADC
52
Minimum ADC in Grade 3-4 gliomas predicts survival
Murakami, R. et al. Radiology 2007243493-499
53
Minimum ADC in Grade 3-4 gliomas predicts survival
Murakami, R. et al. Radiology 2007243493-499
54
Early treatment response reflected by ?ADC
?ADC
Moffat et al, PNAS USA 2005
55
Anisotropy Has High Sensitivity
FA identifies occult transcallosal glioma
infiltration
Stieltjes et al. Neuroimage 2006
56
Anisotropy Has High Sensitivity
FA (better than ADC) predicts tumor infiltration
Stadlbauer et al. Radiology 2006
57
Anisotropy Has Poor Specificity
58
Anisotropy is Pathologically Non-Specific
All of these reduce anisotropy
59
Anisotropy is Pathologically Non-Specific
Tumor or Edema?
60
Anisotropy is Pathologically Non-Specific
Tumor or Edema?
18 gliomas (WHO III-IV), 7 meningiomas, 5
mets ROIs in peritumoral T2 changes
Van Westen et al. Acta Radiol 2006
61
Can directionally-encoded color maps discriminate
edema from tumor infiltration?
62
Vasogenic Edema
Anisotropy Decreased Location / Direction
Normal
Jellison BJ et al, AJNR Am J Neuroradiol 2004
63
Infiltrating Glioma
Anisotropy Decreased Location /
Organization Abnormal
Jellison BJ et al, AJNR Am J Neuroradiol 2004
64
Infiltrating Oligo(astro)cytoma (WHO Grade II)
T2 FLAIR Directional Fiber
Tracking
65
Infiltrating Oligoastrocytoma (WHO Grade II)
66
Infiltrating Oligoastrocytoma (WHO Grade II)
fMRI Brocas Directional DTI
67
Infiltrating Oligoastrocytoma (WHO Grade II)
fMRI Brocas Fiber Tracking
Superior Longitudinal Fasciculus
68
Can directionally-encoded color maps discriminate
edema from tumor infiltration?
No!
69
Applications of Diffusion and Diffusion
Tensor Imaging to Brain Tumors

• Tissue Characterization (DWI / DTI)
• Discriminate tumor from non-neoplastic lesions
• Estimate tumor histology and grade
• Monitor treatment response
• Discriminate edema from tumor infiltration (?)
• Localization / Mapping (DTI)
• Localize lesions to specific WM tracts
• Map WM tracts for pre-op planning intra-op
  guidance
• Assess integrity / relocation of WM tracts post-op

70
Lesion Localization
71
Ganglioglioma
Preoperative Planning
72
Preoperative Planning
PRE-OP
CST
73
Preoperative Planning
PRE-OP
POST-OP
Lazar et al, ISMRM 2004
74
Preoperative Planning
Pilocytic Astrocytoma
PRE-OP
POST-OP
75
Brainstem Glioma
76
Brainstem Glioma
77
Intraoperative Guidance
DTI co-registered with 3D anatomical imaging in a
neurosurgical navigation system
Okada et al, Radiology 2006
78
Nimsky et al, Neuroimage 2006
79
Tractography and Intraoperative Subcortical
Stimulation
Kinoshita et al, Neuroimage 2005
80
Comparison of DTI Fiber Tracking to Subcortical
Motor Stimulation

• Subcortical stimulation sites are
  stereotactically identified on MR using a
  surgical navigation system (Medtronic,
  Broomfield, CO).
• Distance between stimulation site and DTI fiber
  tracks is measured.

Arm Motor Subcortical Stimulation
Average distance of 8.7 3.1 mm from subcortical
stimulation sites to DTI fiber tracks generated
pre-surgically
81
Intraoperative DTI reveals shifting of tracts
during tumor debulking
Nimsky et al, Radiology 2005
82
Proximity of fiber tracts to infiltrating tumors
a function of FA threshold
Stadlbauer et al. Neuroimage 2007
83
Take Home Points

• ADC correlates inversely with cellularity
  (but ROI-dependent in
  heterogeneous tumors!)
• Positive treatment response suggested by early
  ?ADC
• ADC may discriminate high / low grade
• FA decreases with any WM infiltration (tumor or
  edema) and this critically affects tractography
  results through stopping thresholds false
  negatives or decreased reliability of direction
  estimates (false positives)
• Directional color maps and tractography do NOT
  reliably discriminate infiltrating tumor from
  edema
• Tractography helpful for preop / intraop guidance
  but there are pitfalls and validation still
  needed

84
Acknowledgements
UW-Madison Andrew Alexander Brian Jellison
Brian Laundre Mariana Lazar Yu-Chien Wu
Jeffrey Berman Alberto Bizzi Andrei Holodny
Meng Law Pratik Mukherjee John Ulmer
THANK YOU