0% found this document useful (0 votes)
18 views21 pages

MIR2012 Lec4

The document discusses motion artifacts in medical imaging, specifically MRI. It describes how motion during image acquisition can lead to blurring artifacts. Various techniques are presented for estimating and correcting intra-slice motion, including navigator echoes that acquire additional k-space lines to detect motion. The document proposes a new reconstruction method that uses overlapped k-space segments to simultaneously estimate intra-slice motion parameters like rotation and translation. Numerical simulations demonstrate the method can accurately estimate and correct different motion types.

Uploaded by

yeesuen
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
18 views21 pages

MIR2012 Lec4

The document discusses motion artifacts in medical imaging, specifically MRI. It describes how motion during image acquisition can lead to blurring artifacts. Various techniques are presented for estimating and correcting intra-slice motion, including navigator echoes that acquire additional k-space lines to detect motion. The document proposes a new reconstruction method that uses overlapped k-space segments to simultaneously estimate intra-slice motion parameters like rotation and translation. Numerical simulations demonstrate the method can accurately estimate and correct different motion types.

Uploaded by

yeesuen
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 21

Medical Image Reconstruction

Term II – 2012

Topic 4:
Motion Artifacts

Professor Yasser Mostafa Kadah


MRI Data Acquisition
 MR image is acquired in the k-space
 Reconstruction is an inverse Fourier transformation
 Parts of k-space are acquired at different times

-1
Time

k-Space Space
Motion Artifact in MRI
 Motion artifacts result when the patient moves
during MR acquisition
 Physiological/voluntary motion
 Motion artifact manifests itself in the image as severe
blurring that usually mandates the scan to be
repeated
 Costly in addition to added discomfort to the patient
 Postprocessing techniques can be used
 Time consuming and inefficient in many cases
 No considered practical for clinical use
Types of Motion Artifacts
 Intra-slice: motion during acquisition of a slice
 causes k-space of a given image to contain magnitude and
phase errors
 Inter-slice: motion in between acquisition of whole
slices
 causes repeated acquisitions of the same slice to be different
 These two types have been treated separately in the
literature
 Inter-slice motion is simpler to correct for using
registration techniques (e.g., AIR)
Intra-Slice Inter-Slice
Average
Intra-Slice Motion Suppression
 Intra-slice motion artifact suppression is a challenging
problem
 k-space “pieces” are more difficult to register!
 Among the most successful techniques used to
estimate motion is the navigator echo (NAV)
technique.
 Most practical for clinical use.
 The original formulation relies on acquiring an extra
line in the center of k-space along the kx or ky
directions to detect motion in that direction.
Classical Navigator Echo*
 Acquire the navigator (NAV) echo line in the center
of the k-space with every k-space section.
 Each represents the Fourier transform of a projection of the
image
 Register the two NAV lines together to estimate
motion along the NAV direction

Ky=0 Ky=0

* Felmlee and Ehman, Magn. Reson. Med., 1992


Limitations of NAV
 Requires an extra amount of time to acquire this line
prior to actual k-space acquisition
 limits the minimum TE of such sequences
 Additional complexity in sequence programming
 The estimation of motion parameters in both the
read-out and phase encoding directions is not
possible with a single line.
 Two NAV lines in orthogonal directions must be used
 Circular and spherical NAV for 2- and 3-D estimation
Floating Navigator Echo (fNAV)*
 Instead of acquiring the navigator echo line in the
center of the k-space, we acquire this line by
acquiring k-space sections that overlap in a single line.
 Enables the estimation of 2-D translational motion
 Rotation cannot be estimated

* Kadah et al, Magn. Reson. Med., 2004


Arc Navigator Echo (aNAV)*
 A fast way to compute the rotational motion is to
match points on an arc within the area of overlap
rather than the whole area.
 Similar in theory to orbital navigator echo (ONAV)

Arc point (angle)


* Mohamed,Youssef and Kadah, Proc. SPIE Med. Imag. 2003
Reconstruction Method

 Address the problems of intra-slice and inter-slice motion


together
 For example, when segmented acquisition is used with NEX>1
 To propose an extension of the fNAV to allow rotation to be
estimated
 Acquisition of navigator “area” rather than “line” or “arc”
 Take advantage of the extra data acquisition when NEX is required to be
>1 to estimate the intra-slice motion
 Maintain efficiency by not acquiring extra data other than those required
for averaging
Basic Idea
Conventional Acquisition Method
k-space 2 Average
k-space 1
+ =

New Acquisition Method

Acquisition with
overlapped segments
Motion
Estimation &
Correction
Motion-free
Average
Motion Estimation
 Identify the area of overlap under the assumption of a
general in-plane rigid body transformation
 Estimate rotation from magnitude of overlap area
 Correlation based methodology
 Estimate translation from phase of overlap area
 fNAV estimation method
Proposed Method

Overlapped Band Acquisition

Rotation Estimation

fNAV 2-D Translation Estimation

Modified k-Space Values and Sampling Matrix

Gridding

Corrected Image
Experimental Verification Using
Numerical Simulations
 Simulated motion data were obtained from evaluating
the analytical form of the Shepp-Logan phantom with
different motion as well as simulating motion on real
MRI head images.
 Matrix: 128, Band size=16 with 50% overlap.
 Random translational and rotational motion parameters
were simulated for each band
 Reconstruction is performed using conventional
gridding method to account for nonuniformity of
sampling after motion
Simulated Data
 Estimated vs. real motion
Translation Rotation
Distorted Corrected

Distorted Corrected Motion-free


Experimental Verification Using Real
MRI Data
 Real data were obtained from a Siemens Magnetom
Trio 3.0T MR system*
 Matrix 256224
 ETL=16, NEX=2
 Overlap of 50% was used
 Normal human volunteer instructed to move once in the
middle of acquisition
 Reconstruction is performed using conventional
gridding method to account for nonuniformity of
sampling after motion

* Acquired by author at BITC - Emory/Georgia Tech Biomedical Engineering, Atlanta, GA, U.S.A.
Real Data

No Motion Motion
Distorted

Difference
Corrected between
Corrected and
Distorted
Discussion
 Two problems were observed in the reconstruction
phase of the developed method
 Problem 1: Existence of k-space voids
 Missing k-space data
 Undesired variations in the SNR within k-space
 Problem 2: Long reconstruction time
 Rotation requires regridding according to estimated motion
 A new reconstruction table has to be computed each time
Exercise
 Write a short literature review section on the methods used
for inter-slice motion correction in MRI with references.
 Would the proposed method be possible to extend for use
with CT data where acquisition lines are radial? Explain your
answer.
 Use the data set on the class web site to show that 2D
translational motion does not affect the magnitude of k-space
and that such motion can be estimated by correlation based
method.
 Do a literature search on the topic of motion artifacts in ONE
medical imaging modality of your choice and come up with a
list of relevant references related to the subject including both
research papers and patents.

You might also like