Surgery – Diagnostic testing – Detecting nuclear – electromagnetic – or ultrasonic radiation
Reexamination Certificate
1999-05-18
2001-09-18
Casler, Brian L. (Department: 3737)
Surgery
Diagnostic testing
Detecting nuclear, electromagnetic, or ultrasonic radiation
C324S307000, C324S309000
Reexamination Certificate
active
06292683
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates generally to magnetic resonance imaging (MRI), and more particularly to a method and apparatus to track motion, such as anatomy movement, between MR images for efficient and effective MR image registration.
The ability to track motion in a time series of images is essential for a number of different MRI applications. For example, motion artifact suppression techniques require a measurement of motion upon which to take corrective action. Such motion artifact correction has been useful in a variety of applications, including coronary artery imaging such as an MR angiography (MRA), functional MR imaging (fMRI) such as to study brain function, and diffusion imaging. Another motion tracking application is the monitoring of heart wall motion which would be useful to assess the severity and extent of damage in ischemic heart disease. Another application for motion tracking is interventional imaging, such as monitoring the position of a scalpel or other instrument during an interventional procedure.
MR imaging of the coronary arteries, or MR angiography (MRA), has typically been performed using a technique to limit the MRI acquisition to avoid motion artifacts. Such techniques include requiring the patient to withhold breathing during the imaging, using oblique single-sliced image techniques, or respiratory-gated 3D imaging techniques. However, repeated breath holding may not be feasible for many coronary patients and navigation techniques to-date have not generally provided a robust method which works over a range of different breathing patterns in a variety of patients. Another drawback to these approaches is that success or failure is usually not apparent for some time after the start of imaging, and many times not until the imaging has been completed.
Another application requiring accurate compensation for anatomy movement includes myocardial perfusion imaging to detect the passage of a contrast agent through muscle tissue in the heart and to study the blood flow in the micro-circulation of the heart non-invasively. Typically, perfusion imaging consists of using injected contrast agents together with rapid imaging during the first pass with carefully optimized pulse-sequence parameters. Quantification of blood flow from these images is carried out with a region of interest based signal, time-intensity curve analysis. To avoid cardiac motion artifacts, the perfusion images are typically acquired with ECG gating. However, since the period of image acquisition is usually 1-2 minutes long, the images suffer from significant respiratory motion artifacts. This then requires a manual registration and analysis of the perfusion images which is cumbersome and time-consuming because the user must carefully arrange each image to compensate for the respiratory motion before proceeding to a region of interest time-intensity analysis.
The goal of myocardial perfusion imaging is to detect and characterize the abnormal distribution of myocardial blood flow. The ability to extract quantitative perfusion indices such as time-to-peak, contrast enhancement ratio, and the slope from first-pass contrast-enhanced MR images requires a generation of myocardial and blood-pool time-intensity curves for desired regions-of-interest. The computation of these curves is complicated when patients do not suspend respiration adequately, which then results in an image mis-registration over time. This mis-registration occurs frequently due to the fact that the breath-hold duration required to capture first-pass kinetics is typically 20-30 seconds. An accurate spatial alignment of images over a period of time is necessary for creating representative time-intensity curves. Therefore, it would be desirable to have an automatic registration system to track motion and provide automatic compensation for in-plane translation.
A prior art method of MR motion tracking that has had some success is a pattern matching technique. Under this method, an initial region, or pattern, containing the anatomy of interest is saved as a referenced region. In order to track motion occurring in a series of images, the pattern matching technique finds the location in each of the MR images that best matches the initial pattern. The difference between the various pattern matching techniques is the manner in which the best location is chosen. In MR, the previous techniques used for pattern matching have been a least squares technique and a cross-correlation technique. The least squares technique attempts to find a location (&xgr;,&eegr;) in an image f(x,y), that minimizes the squared distance d
2
between the image and the MxN pattern h(x,y) given by:
d
2
(&xgr;,&eegr;)=&Sgr;
i=1
M
&Sgr;
j=1
N
{f(x
i
−&xgr;,y
j
−&eegr;)−h(x
i
,y
j
)}
2
(1)
=&Sgr;
i=1
M
&Sgr;
j=1
N
{f(x
i
−&xgr;,y
j
−&eegr;)
2
−2f(x
i
−&xgr;,y
j
−&eegr;)h(x
i
,y
j
)+h(x
i
,y
j
)
2
} (2)
where (x,y) and (&xgr;,&eegr;) are position coordinates, f(x,y) is the image, h(x,y) is the pattern for matching, M and N are the x,y dimensions of the pattern, and i, j are the summation indices.
The cross-correlation technique is a derivative of the least squares technique and attempts to find the pattern match by maximizing the second term of Eqn. 2. The advantage of cross-correlation over least squares is that it can be calculated very rapidly by Fourier transform techniques. The cross-correlation technique has heretofore been the most widely used method of motion tracking and correction due to the rapid calculation time.
One disadvantage of the least squares and the cross-correlation techniques is that both fail to accurately match patterns in images where there is a local and/or global bias (v) or gain (u) in the image. That is, if f(x,y)=uf
0
(x,y)+v, where f
0
is the initial image and f is the current image. An example of where such bias or gain variations can occur is in the aforementioned case of myocardial perfusion imaging where blood flow into and out of the image plane is expected and causes signal intensity variations. Under such conditions, the least squared and cross-correlation technique may not accurately match patterns.
As can be drawn from Eqn. 2, in order to accurately find the minimum distance between the image and the pattern, cross-correlation requires the additional assumption of constant energy in every M×N region of the image. In practice, however, such an assumption is often not satisfied. This problem is more acute in smaller patterns, such as a 16×16 pattern, than in larger images, such as in 32×32 pattern. As a result, cross-correlation has primarily been used in one-dimensional navigator echoes for the purpose of reducing motion artifacts. Such applications have been largely successful because the cross-correlation technique was used to detect gross motions of the whole anatomy. In these cases, the patterns to be matched were relatively large, and usually, a one-dimensional projection of the whole anatomy. However, for detecting more localized motion in two dimensional images, such as regions of the heart, coronary arteries, or surgical instruments, the cross-correlation technique is ineffective.
It would therefore be desirable to have a method and apparatus capable of accurately tracking motion between MR images that is independent of signal intensity variations, such as local bias, global bias, and/or gain within MR images and/or between MR images to allow either efficient tracking of the motion or for the sake of tracking motion or for using the motion tracking to compensate for motion between images by automatically aligning the images based on the motion tracking.
SUMMARY OF THE INVENTION
The present invention relates to a system and method to track movement in MR images that solves the aforementioned problems. In one application, an automatic image registration algorithm is proposed which greatly reduces the time required for anal
Gupta Sandeep Narendra
Sussman Marshall
Wright Graham
Cabou Christian G.
Casler Brian L.
General Electric Company
Price Phyllis Y.
Ziolkowski Timothy J.
LandOfFree
Method and apparatus for tracking motion in MR images does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for tracking motion in MR images, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for tracking motion in MR images will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2512874