Research - Cardiac MR

 

PERFUSION PROCESSING: CORRECTING PHASED-ARRAY COILS INTENSITY MODULATIONS

James E. Siebert, Radiology; Mark C. DeLano, Radiology;
Joel D. Eisenberg, Cardiology; Jason A. Gift, Radiology

Gift, Siebert, DeLano, Eisenberg

 

Purpose and Introduction

Cardiac MR perfusion studies require the correction of signal intensity variations arising from surface coil reception for both qualitative interpretation and parametric post processing. This work presents an effective correction method suitable for the cardiac application.

Derived parametric images can portray the dynamics information contained in the cardiac perfusion image set (Fig. 1) to aid exam interpretation and to summarize exam results (Fig. 2). Calculation of perfusion-dependent parameters requires a correction of the spatial SI modulations arising from cardiac phased-array surface coil reception. Also, the qualitative appearance of perfusion defects can be dramatically altered by the surface coil sensitivity roll-off (Fig. 3). This work presents an effective correction method suitable for the cardiac application.

We propose an intensity correction method based upon the perfusion image set itself that:

• Requires no additional acquisitions or scan time; ^1-3

• Introduces no additional noise sources; and ^1-2

• Avoids the artifacts of corrections based on low-pass filtering (edge effects, attenuation of anatomic low spatial frequencies). ^1-2

 

Methods

We developed an intensity correction for surface-coil-dependent SI modulation that derives the correction from the perfusion images of the perfusion exam. Specifically, the baseline images of the perfusion acquisition serve as the basis for the intensity correction estimate.  Baseline images are acquired just prior to the Gd arrival into the LV.

Procedure of the Intensity Correction:

1. Detect the first-pass arrival time of the Gd bolus in the LV. This is the end of the baseline period.

2. Average all baseline images together at each slice location. This average slice image provides the basis for the estimated intensity correction.

3. Identify myocardium voxels to serve as a sparse 2D sampling of the smoothly-varying Coil Sensitivity(x,y). Low-pass spatial filtering is applied to further reduce noise error. (Fig. 4)

4. Fit a Minimum Curvature Surface to this sampling. This 2D spline surface serves as the estimate of coil sensitivity(x,y) across entire image area. If the curved-surface fit result contains large excursions, a best fit planar surface is substituted. (Fig. 5)

5. Compute an Intensity Correction Factor(x,y) as a scaled reciprocal of the coil sensitivity(x,y), which is then multiplied pixel-wise with the original image to flatten the image SI scaling. Scaling is set to maintain the same average value of myocardium voxels so viewing display window level is unchanged for the intensity-corrected images. (Fig. 6)

 

Results (Fig. 7)

The performance of the intensity correction method is demonstrated by scanning a body-sized doped-water phantom using the cardiac phased array coil set and the retrospectively-gated FastCard with echo-planar readout perfusion pulse sequence⁴ (Fig. 8). Selecting a contour of voxels in the water can simulate the sampling of LV myocardium. The resulting intensity-corrected image shows that the minimum curvature estimation performs satisfactorily across the image field while assuring a high quality correction in the region of the sampling contour (LV myocardium in actual perfusion images). See also the clinical examples in Fig. 3 and Fig. 7.

 

Discussion

The actual surface coil sensitivity across any image FOV will be always smoothly varying, having at most one trend reversal across the LV. Analytical solution is hampered by undetermined coil shapes (flexible devices), undetermined spatial orientations, variable relative positions for every slice position, and impracticality of additional calibration acquisitions. Thus we attempt this approach to correction estimation based upon minimum curvature surface fitting. In the resulting images, SI scaling is flattened assuredly across LV myocardium, while estimation errors can worsen with extrapolation into distant image regions (irrelevant to perfusion post processing).

In low SI regions distant from coils, lower SNR can become very apparent in intensity-corrected images. The presented method of correcting spatial SI modulation by surface coil sensitivity variations enables the calculation of the first-pass leading SI-increase slope (time rate) image, any parameter dependent on slope measure, and tracer kinetic modeling. Intensity correction also facilitates and simplifies visual interpretation, and enables more effective filming (display windowing).

 

Acknowledgement

Work supported in part by GE Medical Systems.

 

References 

1. Murakami, JW;  Hayes, CE;    Weinberger, E. Intensity correction of phased-array surface coil images. Magn. Reson. Med. 35(4): 585-590 (1996).

2. Grant, PE; Vigneron, DB; Barkovich, AJ. High-resolution imaging of the brain. MRI Clinics N Am. 6:1:139-154 (1998).

3. Thulborn,KR;  Boada,FE; Shen, GX; Christenson, JD;  Reese, TG. Correction of B₁ inhomogeneities using echo-planar imaging of water. Magn. Reson. Med. 39:369-375 (1998).

4. Ding,S; Wolff, SD;  Epstein, FH. Improved coverage in dynamic contrast-enhanced cardiac MRI using interleaved gradient-echo EPI. Magn. Reson. Med. 39(4): 514-519 (1998).

 

Patient Services | Department | Anatomy | Education | Residency | Research | News & Events | MSU | CHM | COM