Research - Cardiac MR

 

MYOCARDIAL PERFUSION RESERVE INDEX IMAGING VIA FIRST-PASS GD DYNAMICS IN VASODILATED AND REST CONDITIONS

James E. Siebert¹, Mark C. DeLano¹, Joel D. Eisenberg^1,2, Sandeep N. Gupta³
Departments of Radiology¹ & Cardiology²
Michigan State University, East Lansing, MI
GE Medical Systems³, Baltimore, MD, USA

 

PURPOSE

To demonstrate the feasibility and benefits of clinical semiquantitative myocardial perfusion reserve (MPR) imaging of the LV myocardium at the native resolution of the MR acquisition for the work-up of ischemic heart disease.

 

INTRODUCTION

The diagnostic accuracy of the myocardial perfusion MR exam remains at 80%-90% when exam interpretation is based upon qualitative evaluation.¹ Since this is roughly equivalent to nuclear medicine, adoption of the MR perfusion exam is not widespread. Myocardial perfusion reserve quantifies the capacity of the circulatory response to a maximal increase in physiological demand.² By providing quantitative objective information to reduce observer variability in MR perfusion exam interpretation, MPR imaging may provide the means to boost diagnostic accuracy, and to document MR myocardial perfusion exam results.

 

METHODS

MR images were acquired every-other heart beat (efgret, 6.6/1.5/25°, 1282, 185 TSR, 8mm, 125 kHz, breath-hold) during the first pass of Gd-contrast antecubital-vein injection (0.05 mmol/kg gadoteridol + 25 ml saline @ 5 ml/sec) in both adenosine vasodilated `stress´, and `rest´ states for 6-7 short-axis locations spanning the LV (GE Signa CV/I 1.5T). ^3-6 The stress and rest dynamic image sequences were post processed to calculate the estimated MPR image for each slice location (~6000 lines of IDL code).

Computation Procedure Applied for Each Slice Location

1. Spatially register across each individual time series at each slice location: Solid-body registration partially compensates for diaphragm motion and cardiac phase jitter.⁸ Usually not required for successful breath holds.

2. Determine a surface coil intensity correction for each slice location: Images acquired before the Gd bolus arrival are analyzed to estimate a surface coil intensity correction at each slice location. The baseline myocardium serves as the reference.⁷ Either a minimum curvature surface or a best-fit plane can be fitted to the myocardial sample points. The scaled inverse of this function is the estimated intensity correction factor. See Fig. 1 & Fig. 2.

3. Calculate first-pass peak upslope parametric images: The first-pass peak upslope event is detected at each pixel x,y to create the Peak Upslope parametric images.⁹ Fig. 3 shows a typical example of the original stress and rest images at Image Index 10, the resulting Parametric Peak Upslope images, and a plot of the time data for this slice location in stress and rest.

4. Perform warping image registration of rest slope image onto stress slope image: Different heart rates during the stress and resting states results in a cardiac phase shift between the 2 sets of registered images at the same slice location--evident in Fig. 4. For pixel-wise processing of the 2 image series, the Rest Upslope parametric image is warp registered onto the Stress Upslope image. The user clicks on corresponding landmark pairs for the LV myocardium in stress and rest images (colored x's).

5. Segment the LV blood pools: Two image regions are determined: the LV blood pool mask (Fig. 5 red areas), the ROI for calculating the average blood pool input function calculation (Fig. 5 blue areas; ~50% of most-enhancing blood-pool pixels). Segmentation can be based upon peak intensities achieved and enhancement time.

6. Calculate the MPR image
   Calculate the MPR(x,y) image pixelwise:
   
   
   
   where SMYOstress(x,y) = first-pass myocardial upslope in stress (green dotted lines Fig. 6), k_Lstress/rest= lumped constants, ∫BP_stress/rest(t) = integral of LV blood pool ROI for input function (see filled red areas, Fig. 6). This expression as written is a corrected stress slope divided by the corrected rest slope. The ∫BP_stress/rest(t) is proportional to the delivered Gd producing the linear part of the increasing SI response. The integration interval is set ≈ the stress LV ROI rise time to peak. The k_Lstress/rest / k_Lstress/rest ratio is presumed = 1.

 

 

Development Clinical Series

Patient volunteers were recruited in outpatient setting: N=15, all having chronic vascular disease. Approximately ≥20 min delay time elapsed between the stress and rest acquisitions. MR perfusion exam interpretation was based on the MPR images (thresholded), with interactive investigation of any suspect regions using interactive time-intensity plot display with crosshair spatial correlation displayed on the original, peak upslope, and computed MPR images. Crtiteria for ‘abnormal’ MPR finding: Segmentally organized defects on at least 2 slices in the thresholded MPR images, unless very convincing single slice defect. The “gold standard” reference was contemporaneous NM and cardiac catheter exams. Exam scoring performed by cardiac MR experienced cardiologist and radiologist.

 

RESULTS

Fig. 7 shows a resulting MPR images for patient having history of old infarct (true-positive patient JAE in Table 1). Normal myocardium shows excellent vasodilation-induced perfusion increase (upper plot), while 4mm away, the defect shows low perfusion in both stress and rest states. Fig.8 histogram shows that values within normal myocardium range from 1.5–5.0, about the expected MPR values. MPR values within perfusion defects range from 0.0–~1.0. Distributions of normal and abnormal myocardium MPR values are separated suggesting that this simplified index approach can be useful for clinical discrimination. (Table 1 - PATIENT EXAMS INTERPRETATIONS AND SCORING)

 

DISCUSSION

Perfusion-dominated SI(t) dynamics occur in the first few seconds of the rising edge of the capillary [Gd]. The rising [Gd] gradient lumen-to-interstitial volumes forces the exchange, and myocardial [Gd] rises with a rate observed as the SI(t) upslope in myocardial voxels. The integral of the blood-pool [Gd](t) will be proportionally related to the delivered mass of the Gd forcing the regional change in observed myocardial SI, with the local perfusive flow modulating the actual local Gd mass delivered. To the extent that lumped scaling phenomena cancel, the MPR(x,y) calculation will reflect the physiologic values. Normal myocardium voxels have calculated MPR values in the range ~1.5-5. The ratio of the integrals serves to correct for any input function variability (cardiac output, bolus dispersion, residual [Gd] from prior bolus). Sensitivity analysis of the integral limits of Eq. 1 shows that stable MPR values result with limits timed per the Stress SI(t) peak (Fig. 1; In the time-intensity plots, the bright SI(t) line is stress data, gray line plots rest data).

Three morphological issues challenge the computation of MPR parametric images:
(1) Cardiac phase shifts between the stress and rest acquisitions arise from the difference in heart rate during the two test states. These morphological differences necessitate a warping image registration. (2) Stress-rest difference in diaphragm positions cause mismatch in ventricular anatomy that must be compensated via warping image registration. (3) Cardiac phase jitter within a given slice location image sequence results in variable partial volume averaging of the myocardial edge voxels, which introduces variability over time in LV edge features. Stress-rest mismatches of myocardial anatomy may pose the ultimate limitation of MPR imaging.

Intensity correction of surface coil reception modulations is needed to determine the input function normalization (slope spatial dependency), to achieve post processing automation, as well as to improve the qualitative perception of cardiac perfusion images (Fig. 2) , movie loop presentation, and, very importantly, the perception of time-intensity curves during interactive review of MPR quantitative images.

Figure 7 presents typical MPR imaging patient results. Given the integrative and objective MPR images, interactive investigation of the dark suspicious regions in the thresholded MPR image forms the core of perfusion exam interpretation. Qualitative assessment of the MR perfusion images is a very challenging perceptual task. Quantitative analysis may contribute to improving the MR exam diagnostic accuracy.

Our current work is aimed at reducing variability arising from the current surface coil intensity correction implementation. Signal intensity in the efgret images is very low prior to Gd arrival.

The feasibility of high-resolution practical clinical MPR images has been demonstrated. MPR imaging may provide quantitative objective information to reduce variability in perfusion exam interpretation, and to document MR myocardial perfusion exam results.

 

ACKNOWLEDGMENT

Research supported in part by GE Medical Systems.

 

REFERENCES 

1. Taillefer R, DePuey EG, et al. J Am Coll Cardiol 29(1): 69-77, 1997
2. Wilke N, Jerosch-Herold M, et al. Radiology 204:373-384, 1997
3. Ding S, Wolff SD, Epstein FH. Magn Reson Med 39(4):514-519, 1998
4. Slavin GS, Wolff SD, et al. Proc ISMRM 8th Mtg, p 36, 2000
5. Nidal A-S, Nagel E, et al. Circulation 101:1379-1383, 2000
6. Jerosch-Herold M, Wilke N, et al. Proc SMR 3rd Mtg, p 459, 1995
7. Siebert JE, DeLano MC, et al. Proc ISMRM 7th Mtg, p2179, 1999
8. Gupta SN, Foo TK, et al. Proc ISMRM 7th Mtg, p2178, 1999
9. Schwitter, J., Nanz, D., Kneifel, S., et al. Circulation 103, 2230.5, 2001
   

Download presentation in PDF format

 

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