Imaging three-dimensional motion in the heart using zHARP Jerry L. Prince

advertisement
Imaging three-dimensional motion
in the heart using zHARP
Jerry L. Prince
Image Analysis and Communications Laboratory
Electrical and Computer Engineering
Johns Hopkins University
Acknowledgments
•Nael Osman
•Jerome Garot
•Elias Zerhouni
•Elliot McVeigh
•Ergin Atalar
•Andy Derbyshire
•Tom Foo
•Carlos Rochitte
•Alan Heldman
•Li Pan
•Matthias Stuber
•David Bluemke
•Joao Lima
•Ernesto Castillo
•Dara Kraitchman
•Smita Sampath
•Khaled Abd-Elmoniem
•Bernard Gerber
•William Kerwin
•Sandeep Gupta
•Harsh Agarwal
•Vijay Parthasarathy
NIH/NHLBI, Whitaker Foundation, GEMS, Institut Roche
Cardiovasculaire, Federation Francaise de Cadiologie
Outline
•Introduction
•MR Tagging and HARP
•FastHARP
•SF-HARP
•zHARP
•Conclusion
Outline
•Introduction
•MR Tagging and HARP
•FastHARP
•SF-HARP
•zHARP
•Conclusion
Coronary Artery Disease
•Accumulation of plaque
in coronary arteries
•Lack of blood flow and
oxygen Ischemia
•Prolonged Ischemia
Infarction
•Weakening of the heart
 Heart Attack
•Detection &
Quantification of
Ischemic and Infarcted
tissue
Cover page of Scientific American, May 2002
MRI of Cardiac Motion and Strain
•Methods:




Anatomical imaging: CINE MRI
Phase contrast: PC-MRI [Wedeen et al. MRM 1992],
Stimulated echo: DENSE [Aletras et al. JMRI 1999],
Tagging [Zerhouni et al. Rad. 1988, Axel et al.
Rad.1989]
•Myocardial motion and strain are reliable
regional & global diagnostic indicators of
CAD.
Can be used during rest or under stress
(induced using pharmaceutical)
•
Clinical Goal 1
Real-time free-breathing
technique: 2-D motion and
strain
Early detection of ischemia
during MR stress tests
FastHARP
03]
[Sampath et. al. MRM
Clinical Goal 2
Rapid 3-D motion and strain
imaging (1-4 short breathholds)
Diagnosis and treatment
planning of patients with CAD
during resting conditions
SF-HARP
[Sampath et. al. ISMRM 03]
Clinical Goal 3
Dense 3-D motion and true
planar strain
Diagnosis and treatment
planning of heart failure
patients
zHARP
[Abd-Elmoniem et al., ISMRM
’05, SCMR ‘06]
Outline
•Introduction
•MR Tagging and HARP
•FastHARP
•SF-HARP
•zHARP
•Conclusion
GE Scanner System
Four channel phased
array cardiac receiver coil
GE Signa CV/i whole body MR clinical scanner system
1.5T magnet, gradients: 40mT/m, slew rates: 150mT/m-ms
ANSI C in the GE pulse sequence programming environment : EPIC
Cardiac MR Tagging
[Zerhouni et al., 1987, Axel et al., 1988]
•Put noninvasive markers inside the
myocardium; markers move with tissue
•Evaluate regional myocardial function
quantitatively and noninvasively
Initial Time
Later Time
Right
Ventricle
Short-axis
Images
Left
Ventricle
Principle of HARP Image Analysis
Reference time:
sinusoidal tag pattern
Later time:
Tissue compression
increases frequency
Sinusoidal
tag pattern
Computed
phase of
tag pattern
“wrapping” artifact
1. Slope of phase increases
2. Phase values are constant
t=0ms
tagged images
k-space = Fourier Space
t=390ms
Computing a Harmonic Image
Fourier
transform
Inverse
Fourier
Transform
Bandpass
filter
(Complex) Harmonic Image
Real Part
Imaginary Part
Harmonic Magnitude Image
• The harmonic magnitude image D is a blurred MR image
without the tag patterns
• By simply thresholding the magnitude image, a segmentation
l
mask is produced
Magnitude
Mask
Harmonic Phase (HARP) Image
• The harmonic phase image
• can only be computed between
and
• The computed harmonic phase
(HARP) angle image is
• W is a nonlinear wrapping function
• Motion information is *not* limited to
the wrapping artifacts  resolution
is better than conventional tag lines
magnitude mask
2-D Motion
Vertical
Horizontal
Principle of Point Tracking
[Osman, Kerwin, McVeigh, Prince, 1999]
i( yn1 , tn1 )  i( yn , tn )
i=1,2
•Track a pair of HARP values – horizontal and
vertical – throughout an image sequence
y
y
Initial time
y’
y’
Later time
2D HARP Tracking
Track Grid
Computing Lagrangian Strain
• Simple Lagrangian strain is change in length per unit
length.
• Grid provides points for circumferential and radial strain
for sub-endocardium, midwall, and sub-epicardium
• The strain between q and q is
1
2
Lagrangian Circ Strain Profiles
Percent Strain
epi
septum
endo
septum
pacer lead
Stretching >0
pacemech9
Shortening <0
[Data courtesy of
Elliot McVeigh, 1998]
Computing Eulerian Strain
•Slope of the
harmonic phase
gives strain
The local elongation
in the direction of n
is computed using
•
•Define matrix of tag
frequencies
Ω  [ w1 w2 ]
Eulerian Circumferential Strain
Radial
Strain
Circumferential
Strain
Normal Volunteer
Eulerian Circumferential Strain
Normal Volunteer
What is HARP?
•Phase-based optical
flow?
 Fleet and Jepson
(1990)
 In many ways HARP is
simpler than this
•Frequency or phase
demodulation?
 Havlicek, Harding, and
Bovik (2000)
 Standard demodulation
methods are not
adequate
 A new MR imaging
method?


HARP began as a
tagged image
processing method,
but has evolved to
something more
Now, we believe
that
HARP is a new way to
image regional cardiac
function
Resolution and Dynamic Range
•HARP resolution
approximately equal
to Fourier resolution
Fourier Acquisition Box
•HARP dynamic range
determined by Fourier
acquisition “box”
Outline
•Introduction
•MR Tagging and HARP
•FastHARP
•SF-HARP
•zHARP
•Conclusion
HARP Requires Less Info
Inverse
Fourier
Transform
Reduced k-space Acquisitions
Conventional imaging : 16 heartbeat breath-hold
2 heartbeat breath-hold
HB1
HB2
FastHARP Imaging
Protocol
[Sampath et. al.,ISMRM 2001]
[Sampath et. al., MRM 2002]
HB 1
APPLY
TAGS
ACQUIRE IMAGES
32 x 32 k-space acquisition “box”
Temporal resolution = 40ms
FastHARP Imaging Protocol
4 heartbeat
View-Sharing the Acquired Data
ET=1
ET=2
ET=3
ET=4
ET=1
ET=2
ET=3
ET=4
Conventional Grouping t=38.8ms
…
View-shared Grouping
t=9.7ms
…
FastHARP Tracking
Normal volunteer
Yellow lines:
Tracked trajectories
Blue asterisk:
Position of the tracked
material point in the
time frame in view
FastHARP Circumferential Strain
FastHARP Radial Strain
Continuous Monitoring Mode
Continuous acquisition of harmonic images with
alternating tagging directions every successive
heartbeat.
Reference scan is required only once.
V
H
2
1
Strain Maps
V
3
Strain Maps
H
N-2
V
N-1
Strain Maps
H
N
Strain Maps
Ischemic Dog Studies
[ Kraitchman, et. al., Circulation 03]
•A short axis slice was prescribed
•Balloon is inserted into coronary artery
•2-heartbeat reference scan
•Continuous FastHARP acquisition for about 1 minute
•Imaging Parameters:
 32 X 32 matrix
 320 FOV
 RBW: 62.5kHz
 Temporal resolution: 38.8ms
End-Systolic Circumferential Strains
Sequence of events
during imaging:
Event
time (s)
Balloon Up
20
Mild EKG changes
50
Strong EKG changes
75
Balloon Down
120
End-Systolic Circumferential Strain
HB time (s)
14
20
30
50
45
75
65
120
Summary of FastHARP
•FastHARP pulse sequence (multi-shot EPI)
•Single-shot mode, continuous mode
•Real-time (one heartbeat delay)
•Real-time strain computation and display
Main Contributions:
 Real-time, free-breathing
 2-D motion and strain imaging protocol
 Detects onset of ischemia during MR
dobutamine stress tests
Outline
•Introduction
•MR Tagging and HARP
•FastHARP
•SF-HARP
•zHARP
•Conclusion
SF-HARP Goals
•Use FastHARP in 3-D
imaging protocol to
reduce number of
breath-holds
Use HARP in
postprocessing to
reduce analysis time
Make fully automatic
and model-free
•
•
Conventional 3-D Tagged MRI
HARP Tracking
[Osman and Prince, IEEE Transactions in Med. Im. 2000]
Through plane motions of material points
out of the fixed image slice are not taken
into consideration.
CSPAMM
•Image twice, cosine tag then –cosine tag
•Complex subtraction
time
Slice Following with CSPAMM
[Fischer et. al., MRM 1994]
Imaged slabs
tm
Tagged Slice
tm+1
Signal from
tagged slice
A(p, t )  D(p, t ) cos[ (p, t )]  D 0(p, t )
Signal from
untagged
regions in the
imaged slab
B(p, t )  D(p, t ) cos[ (p, t )   ]  D0 (p, t )
I SF CSPAMM (p, t )  A  B  2 D(p, t ) cos[ (p, t )]
SF-HARP images (Short Axis)
A
B
F.T.(A-B)
A
F.T.(A-B)
B
True 2-D Tracking (SA slice)
Tagged Slice at tm+1
pm+1
p'm+1
pm
p'm
X
Image Slab
Tagged Slice at tm
SF-HARP Images (Long Axis)
A
B
F.T.(A-B)
A
F.T.(A-B)
B
True 2-D Tracking (LA slice)
Tagged Slice at tm+1
pm+1 p'm+1
Imaged slice
pm p'm
Y
Tagged slice at tm
True 3-D Tracking Method
Point at intersection of LA
and SA image planes
LA
SA
SF-HARP Pulse Sequence
Slice Following and CSPAMM
features
Six heartbeat acquisition per slice.
HB1&2: Reference dc images
A
A
HB3&4: HARP images with (90,90)
SPAMM tags (A)
HB5&6: HARP images with (90,-90)
SPAMM tags (B)
B
B
Experiment
•Eight SA slices – 2 short breath-holds of 20s each
•Six LA slices – 2 short breath-holds of 15s each
•Sequence of 12 SF-HARP images obtained over 85%
of R_R.
•Imaging Parameters:
32 X 32 matrix
320 FOV
rbw 62.5KHz
imaging flip angle of 20
temporal resolution=48ms
Data Analysis
•Lines of intersections
were determined
•A regular grid of
points on these lines
was defined.
•For each slice, 2-D HARP
tracking was performed
on the grid of points
•Trajectories were
combined to obtain the
true 3-D motion.
BASE
Results (Top View)
APEX
Apical slice
Mid-ventricular slice
Notes: 1. Anticlockwise rotation during systole
2. Clockwise rotation during diastole
Basal slice
Average Rotation
Note: 1. Apical to Basal Twist
Results (Front View)
Septal
Free-wall
Anterior
Posterior
Notes: 1) Basal slices push downwards
2) Apical slices push slightly upwards
free-wall
septum
3) Longitudinal compression, Radial Thickening
4) Increased compression in the free-wall
posterior
anterior
Average Longitudinal Compression
Anterior
Posterior
Circumferential and
Radial Strains
anterior
free-wall
septum
posterior
Summary of SF-HARP
•3-D tracking of material points
•Imaging in 4 short breath-holds
•Strains and torsions can be computed from the
tracked markers
Main Contributions:
• Completely data driven (no model).
• Fast image acquisition.
• Post-processing completely automated
• Global and regional diagnostic indicators
Outline
•Introduction
•MR Tagging and HARP
•FastHARP
•SF-HARP
•zHARP
•Conclusion
Rethink Slice Following
• HARP tells us p goes to q on the image plane
• It does not tell us the z-component of displacement
• Let us use a z-phase encode to “label” the z-component
zHARP Pulse Sequence
90o
±90o
RF
zGradient (z-encode)
+Az
A: vertical
tagging
Gx
B: horizontal
tagging
CRUSHER
Gz
-Az
Gy
Slice
Selection
With SF-CSPAMM
Acquisition
With zGradient
Decoding the Phases
B
FT
FT
C
D
HARP Filtering
& Processing
A
Solving
Ax=b
zHARP Process
•Compute (Eulerian) phases:
•Procedure:
 Use
and
to find apparent 2D motion

uniquely determines z position
•All points on image plane can be
uniquely tracked in 3D
Experiment 1: (Through-plane motion)
Simple Through-plane Motion
Through-plane & In-plane motion
zHARP Human Subject
SEPTAL
ANTERIOR
LATERAL
INFERIOR
z Displacement tracking results for points around the
myocardium. Vertical axes are in mm and horizontal
axes are the time-frame index.
False In-plane Strain
L
Stripe at
time=t0
φ
Stripe at
time=t1
Throughplane
(z)
L
Apparent
Image at
t10
L0=L
In-plane
(x)
Displacement (u)
y
zHARP can
correct for false
2D strain
L1=L cos(φ)
x
uy(t1)
ux(t1)
uz(t1)
with
conventional
tagging
Needed
Strain Mapping
Displacement
gradient
With conventional tagging
= 0 (No rotation
through-plane)
= 0 (No compression
within-slice)
= 0 (No shear
within-slice)
Using through-plane
motion from zHARP:
LA
SA
SA
LA
Phantom: a jar
filled with
electrode gel.
zHARP Tracking of SA and LA
Acquisition Window: 10ms
Spiral Interleaves: 20
Res. 256x256,FOV 320mm
TE 1.1ms, TR 30ms
Slice Thickness 8mm
Tag-spacing 8mm
before
correction
4 Eulerian strain maps at sample time frames
LA
1
3
2
Strain %
after
correction
using
zHARP
Regional Eulerian Strain
vs. time
Before correction
After correction
2 1 87
3
45 6
•Acquisition Window: 15ms,
•12 spiral interleaves, TE/TR 4.0ms/30ms
•Res. 256x256,FOV 350mm,
•Slice Thickness 6mm, Tag-spacing 8mm
% Circumferential Strain
Before correction
After correction
% Radial Strain
Surface Strain in Normal Heart
Summary of zHARP
•Dense 3-D tracking of material points
•Imaging a plane in one breath-hold
•Combines HARP and phase contrast concepts
Main Contributions:
• Completely data driven (no model)
• Fast image acquisition
• Post-processing completely automated
• Corrects rotational strain artifact
Outline
•Introduction
•MR Tagging and HARP
•FastHARP
•SF-HARP
•zHARP
•Conclusion
Conclusion
•HARP is now extended to 3D
 SF-HARP for global measures in a few
breath-holds
 zHARP for dense motion on a plane and
true planar strain
•Future work:
 improve imaging speed
 automate myocardial segmentation
 extend zHARP for true 3D strain
The End
Conflict of Interest Disclosure
Jerry L. Prince is a founder of and
owns stock in Diagnosoft, Inc., a
company which seeks to license the
HARP technology. The terms of this
arrangement are being managed by
the Johns Hopkins University in
accordance with its conflict of
interest policies.
Download