MRA Acquisition & Analysis

Acquisition and Analysis of Magnetic Resonance Angiographic Imaging Studies

AIO Network Netherlands
Jos J.M. Westenberg, Ph.D.

Background

In this project, MRA studies were acquired and analyzed. Special attention was given to renal and peripheral arteries.

Goals

Stenosis quantification methods were proposed and exercised. The results were compared to the results from conventional imaging modalities, such as X-ray angiography.

Approach

The following subjects were addressed during this study:

Objective and semi-automated stenosis quantification from post-stenotic signal loss in 3D phase-contrast MRA of flow phantoms and renal arteries.

3D Phase-Contrast (PC) MRA is an MR technique which displays the moving spins inside a volume (i.e., the flowing blood in arteries). A stenosis in an artery is characterized by signal loss distally to the stenosis. This post-stenotic signal loss (PSL) occurs due to the complex blood flow pattern that exists at the site and distally to the stenosis. Inside an MR measurement volume (a voxel), several velocity values will be present, yielding phase dispersion. This is a known cause for signal loss. Also, during the time elapsed between exciting the spins and recording the echo signal (i.e., the TE), the spins inside a voxel will be mixed with other spins due to the complex flow, yielding further loss in signal.
In Figure 1-1A, the maximum intensity projection (MIP) of a flow phantom (made out of heat shrink tube) is presented, which is recorded with 3D PC MRA. In the flow phantom, a stenosis is applied. The reference diameter amounts to 6.80 mm, the obstruction diameter is 2.92 mm. This yields a diameter stenosis (%D) of 57%. The phantom is included in a stationary flow circuit, with a coppersulphate solution in water as a circulating fluid (T1 = 777 ms, T2 = 266 ms). A flow of 440 ml/min is applied. Along the axis of the MIP of this phantom, a measurement line is drawn, and the signal intensity (SI) along this line is plotted in Figure 1-1B.

SI in PC MRA is proportional with the velocity of the spins measured in the image. Due to saturation effects, especially in 3D imaging where spins flow in a volume which is continuously excited, the SI will decrease, even if the velocity is constant. This effect is exponentially, but from our experiments it is proven that over a small trajectory this can be approximated as a linear decrease. PSL is noticeable by the sudden drop in SI, which gradually recovers downstream.
In the literature, stenosis severity is linked to the length of the PSL. This length is usually measured on the viewing console. This method introduces several errors: viewing parameters (i.e., window level and width) will effect the definition of the start en endpoint of the PSL, as will inter-observer variabilities. Furthermore, the actual amount of signal loss (which is related to the severity of the stenosis) is left out of consideration by measuring solely the length of the area of signal loss.
Therefore, we have introduced a new, observer independent and semi-automated method for quantifying the amount of signal loss (defined as the severity of PSL) from the SI-plots as presented in Figure 1-1B. This method is illustrated in Figure 1-1C. The SI is modeled by a linear fit on the undisturbed part of the SI-course (i.e., where the saturation has a decreasing effect on the SI). A deviation from this modeled course due to PSL is detected automatically and the region of PSL is indicated. The length of this region is measured with no observer-dependency, nor do parameters on the viewing console have a role in this procedure. The severity of the PSL is defined as the ratio of the residual area in the SI-plot (i.e., the SI which is actually lost due to the stenosis) and the area of SI which is found under the modeled SI-course for the same area. This severity of PSL (in %) can be determined for each stenosis in MIPs, and compared to the value of the length of the PSL.

For three stenotic phantoms (i.e., with diameter stenosis values 31%, 57% and 71% respectively), the length and the severity of signal loss are determined for five flow values, ranging from 300 ml/min to 800 ml/min.The results are presented in Figure 1-2.

The severity of PSL correlates well with %D stenosis. For a larger stenosis, a higher value of severity of PSL is expected. Also, the severity of PSL correlates with the flow value, since higher flow values yield a larger area of complex flow distal to the stenosis and thus more signal loss.
The length of PSL does not seem to correlate as well with %D stenosis as the severity of PSL.
These in vitro experiments indicate that the severity of PSL might be a useful parameter for stenosis quantification in MIPs of 3D PC MRA datasets. This is now tested in vivo. Ten patients with suspected renal arteries were selected, yielding 17 patent main arteries (3 occluded, confirmed with X-ray DSA). On the DSAs, four hemodynamically significant orifice stenoses were found , six hemodynamically significant non-orifice stenosis and one artery showed a strong curvature at the orifice, but no stenosis was found. Six arteries showed no stenosis on the DSA.
Seven healthy volunteers were selected, yielding 14 arteries. SI-analysis was performed on the MIPs, oriented in a view corresponding with the DSA. SI-plots were divided into four typical patterns. Examples are shown in Figures 1-3, 1-4 and 1-5.

All arteries with a significant orifice stenosis (4) showed a similar SI-plot as shown in Figure 1-4A. This type of SI-plots can not be analyzed with the PSL quantification algorithm as explained above, but this typical SI-course was found consistently for all these stenoses. All arteries with a non-orifice stenosis (6) showed an SI-plot similar to Figure 1-4C. All arteries of the healthy volunteers showed an almost linear decreasing SI similar to Figure 1-4D.
The PSL was quantified and the length and severity were determined. The relationships between these parameters and the %D stenosis determined with DSA are presented in Figure 1-6. In these scatter plots, data of two more arteries were included which showed no stenosis on the DSA, but the PSL quantification showed a small region of signal loss.

No significant correlation was found between the length of the PSL and the %D stenosis (r = 0.37, p = 0.37), but the correlation between the severity of the PSL and the %D stenosis was statistically significant (r = 0.90, p = 0.002).
This method presents a semi-automated and observer-independent way of quantifying PSL. The severity of PSL correlates better with the stenosis severity than the length of PSL. This parameter incorporates the degree of signal loss due to the significance of a stenosis, rather than just indicating the region of signal loss. For testing the usefulness of this parameter for stenosis quantification, more in vivo data is necessary and an ROC-analysis has to be performed.

Variations in blood flow waveforms in stenotic renal arteries by 2D phase-contrast cine MRI

Blood flow measurements with two dimensional (2D) Phase-Contrast (PC) Cine MRI are performed by placing a measurement plane perpendicular to a blood vessel (see Figure 2-1). The velocity through this measurement plane is proportional to the phase of the magnetization. In the phase image, the gray value is proportional to the velocity value. By indication a region of interest (ROI), the flow through this ROI (i.e., the cross-section of a vessel) can be determined by integrating the velocity over the area of the ROI.

The flow is determined in 22-25 phase images during the cardiac cycle, and the flow in ml/s against the time is plotted in a flow pattern. Three parameters are determined from this flow pattern: the total flow in ml/min, the systolic wave duration Dt in ms, and the diastolic flow value in ml/s. In Figure 2-2, flow patterns are presented of a measurement proximal (A) and distal (B) to a hemodynamically significant stenosis (moderate on DSA, transstenotic pressure drop DP = 150 mm Hg, intra-arterially measured).

The shape of this renal blood flow pattern (i.e., the parameters) is not only determined by the presence and severity of a stenosis, but this varies for each individual subject. Age is an important factor influencing the blood flow. Arteries in older subjects have less distensible wall behavior, resulting in a less damping effect compared to arteries in younger subjects. Older subjects have a lower cardiac output compared to younger subjects, resulting in lower blood flow values.
In this study, distinction is made between older and younger subjects. Seven healthy younger volunteers (i.e., 13 flow measurements) with a mean age of 28 years and five healthy older volunteers (i.e., 10 flow measurements) with a mean age of 58 years were recruited. Fourteen patients (i.e., 15 flow measurements) with a mean age of 60 years were included. For patients, stenosis severity is determined with the intra-arterially measured transstenotic pressure drop. A stenosis with DP > 15 mm Hg is considered hemodynamically significant. The results for four groups of flow measurements (i.e., arteries younger healthy volunteers, arteries older healthy volunteers, non-stenotic arteries patients and stenotic arteries patients) are presented in Table 2-1.

 

Healthy Volunteers

Patients

 

Young
(group 1)
(n = 13)

p-value

Old
(group 2)
(n = 10)

p-value

No Stenosis
(group 3)
(n = 6)

p-value

Stenosis
(group 4)

Total blood flow (ml/min)

603 ± 21

<0.001

454 ± 16

0.91

497 ± 130

0.24

264 ± 52

(n = 9)

Dt (ms)

307 ± 4

0.002

395 ± 27

0.51

366 ± 30

0.03

507 ± 48 (n = 5)

Diastolic flow (ml/s)

7.9 ± 0.4

0.002

5.9 ± 0.2

0.17

4.7 ± 1.4

1.00

3.9 ± 0.9 (n = 5)

Table 2-1: Mean values and standard errors for flow parameters obtained in arteries in younger healthy subjects (group 1), older healthy subject (group 2), non-stenotic arteries in patients (group3) and stenotic arteries (group 4), classified by DP > 15 mm Hg. The p-values show statistically significant (indicated in red) differences between the mean values of particular groups, determined with Mann-Whitney U tests.

Significant differences between flow parameters determined for different groups are determined with Mann Whitney U tests. The p-value shows the statistical significance of this difference. From Table 2-1, significant differences for total flow, systolic wave duration and diastolic flow are found between healthy younger subjects and healthy older subjects. Younger subjects have a higher total and diastolic flow value, as well as a shorter systolic wave duration. No differences were found between the flow parameters determined in the arteries of the older healthy volunteers and the non-stenotic arteries of the patients, which are comparable in age. A significant longer wave duration was found in the stenotic artery, though, compared to non-stenotic arteries.
From this study, it can be concluded that a systolic wave traveling through a stenotic renal artery becomes damped. This effect was already known from echo Doppler studies, but is shown here with 2D cine MR flow measurements. Also, an age-related effect on the renal blood flow pattern was shown with MRI.

Scan optimization of gadolinium contrast-enhanced 3D MRA of peripheral arteries

In this study, a new contrast-enhanced (CE) MR multi-station scan protocol is developed to study the complete arterial systems of the peripheral arteries, from the aorta bifurcation down to the distal run-off. The peripheral arteries are acquired in three adjacent stations, with some overlap, to image the entire area of interest. Per station, a Gadolinium (Gd) contrast agent is injected intravenously (i.v.) as a bolus.
The patient is positioned supine in the MR scanner, and the legs and feet are strapped in a foam cushion to minimize motion artifacts. An 18-gauge Venflon-2 i.v. catheter is placed in a forearm vein and a Gd-dose is injected through a Spectris MR Injector, followed by a saline flush at the same infusion rate. The injector is filled with 45 ml of Gd and 60 ml of saline. After each Gd-injection, a 8-10 ml saline flush is administered, sufficient to flush the Gd through the supplying tube into the vein. The center of the central section is positioned with the light visor at a level just above the knees, at ± 600 mm from the feet. This marks the center of the total imaged volume. Then the center station is imaged first with a scout view.
Next, the patient is moved 400 mm cranially, and the distal station is imaged with the scout view. Then the patient is moved 800 mm caudally, to place the proximal station in the isocenter of the magnet. This station is also imaged with a scout view.
The proximal station is visualized first with 3D Gd CE MRA, then the central station and finally, the distal station. First, the timing of bolus arrival is determined. Therefore, a two-dimensional (2D) T1-weighted gradient-echo sequence is used, with the plane oriented perpendicular to the aorta, just above the bifurcation. 2 ml Gd contrast is infused as a testbolus at an infusion rate of 1 ml/s. Starting at the moment of infusion, the 2D image is acquired every second, during 90 seconds after the infusion. From a time-intensity diagram, the delay-time of signal enhancement (i.e., bolus arrival) in the aorta is determined from measuring the signal intensity in a region-of-interest (ROI) placed in the cross-section of the aorta.
Next, the arteries are imaged as dynamic 3D datasets. A series of two datasets is acquired, one without Gd (i.e., the mask images) and one Gd CE dataset. Image analysis is performed after digital subtraction of these datasets. The proximal station is imaged using a breath-hold (to suppress respiratory motion artifacts) T1-weighted Turbo Field Echo sequence with 25 s imaging time. Centric k-space acquisition is used. 15 ml of Gd is injected at an infusion rate of 1 ml/s.
Delayed image reconstruction is used, so the imaging of the next station can be performed immediately after the previous station. The patient is moved 400 mm cranially and the central station is acquired using a 3D T1-weighted spoiled gradient-echo technique. Linear k-space acquisition is used. Imaging time is 60 seconds, and a bolus of 12 ml Gd is used at an infusion rate of 0.3 ml/s. No additional timing of bolus arrival is performed, taking into account the slower infusion and the extra trajectory for the bolus to travel from the proximal to the central station, 5 s are added to the bolus arrival time to determine the scan delay for the acquisition of the central station. Again, a mask dataset is performed before Gd is infused and the Gd CE dataset is acquired. Subtraction will suppress Gd contrast already present in veins and in the interstitial space.
Finally, the patient is translated 400 mm cranially and the distal station is imaged identically as the central station, now with 16 ml Gd and again no testbolus. The same scan delay as for the central station is chosen.
In Figure 3-1, MIPs of the 3D CE Gd datasets are presented of the proximal (A), central (B) and distal (C) station of a patient with a bi-iliac prosthesis. The entire arterial trajectory is clearly visualized.

In Figure 3-2, a patient with a right iliac occlusion immediately at the bifurcation is presented. In the MIP of the Gd CE MRA dataset (A), patency of the distal vasculature due to collateral filling is clearly visible, while on the X-ray image (B), this is not as well demonstrated. This clinical condition is quite common in patients with peripheral arterial disease. Accurate delineation of the run-off vessels is very important in the decision-making process for treatment of these particular patients.

The scan protocol is optimized using five patients with proven peripheral arterial disease. Three repetitive bolus injections showed sufficient image enhancement for imaging the complete arterial bed of the pelvis and the legs at a dose of 45 ml Gd in three overlapping stations, each with a field-of-view of 450 mm. Venous enhancement and enhancement in the surrounding tissue were adequately suppressed by subtraction. The total examination time was less than 1 hour. The images were of good quality which allowed adequate clinical evaluation and the diagnoses taken from the MRA examinations made by two experienced vascular radiologists were in agreement with the diagnoses based on the X-ray images.

Vessel diameter measurements in gadolinium contrast-enhanced 3D MRA of peripheral arteries

Vessel diameter measurement in angiographic imaging studies is the first step towards stenosis quantification. In this study, Maximum Intensity Projections (MIPs) of Gadolinium (Gd) Contrast-Enhanced (CE) MRA datasets of the peripheral arterial bed are studied. The MRA datasets are acquired with a previous described scan protocol. Vessel diameter measurements are performed in the MIPs with an automated method. Comparison to vessel diameter measurements in X-ray angiographic images is performed only in the proximal station ((i.e., the aorta, bifurcation and the iliac arteries) of the arterial bed.
It is known from the literature, that measurements in MIPs are often underestimated, because signal intensities at the border of the vessels are suppressed by background voxels. This can be solved by using measurements in original acquisition planes, or by limiting the projection space. Using the original imaging planes is not feasible, since the images are acquired coronally. Therefore, a vessel track is constructed of the artery of interest, and around this track, the dataset is clipped; 10 voxels are considered for the selective MIP. An example of this construction is presented in Figure 4-1.
Vessel diameter measurements are performed on both the total MIPs and the selective MIPs and compared to measurements in the X-ray angiograms. The diameter is determined automatically from the signal intensity (SI)-plot constructed from a measurement line drawn perpendicular to the vessel. The maximal SI and the SI in the background are determined, and by using a threshold value, the diameter of the vessel can be determined. This threshold value is expressed in a percentage of the range between the maximal SI and the SI in the background. In Figure 4-2, this algorithm is schematically drawn.
The threshold value depends on the MR technique used. The value is empirically determined by diameter measurements in phantoms with varying threshold values. The phantoms were imaged in previous studies with Phase Contrast (PC) (three phantom experiments with a constant flow of 200 ml/min) and Gd CE MRA (five phantom experiments with a constant flow of 200 ml/min). Of each experiment, two MIPs were analyzed and the mean diameter value and additional standard deviation are determined with the algorithm, for threshold values varying from 10% to 50%. The results are plotted in Figure 4-3.
For PC MRA, the optimal threshold (i.e., where the diameter value reaches the true diameter value) is 10%. This corresponds with results known from the literature. For Gd CE MRA, the optimal threshold value is 30%. This value is now used in the algorithm for diameter measurements.
In vitro, the validity of diameter measurements for low resolution images is tested. Stenotic phantoms with obstruction diameters ranging from 5.58 mm to 1.97 mm are investigated and diameter measurements are performed for both PC MRA as well as Gd CE MRA. The inplane voxel size amounts to 1.75 mm for Gd CE MRA and 1.87 mm for PC MRA. The % Error in diameter measurement is determined (with the phantom diameter measurement determined with X-ray as gold standard) and plotted on a log-scale versus the gold standard in Figure 4-4.
The percent error increases approximately exponentially with decreasing diameter for both MRA techniques, yielding a large overestimation for small diameters. For a diameter value of 4.69 (2.5 pixels per diameter), the percent error reaches an acceptable value of 9% overestimation for PC MRA. The percent error equals 4% for a diameter value of 5.58 mm (3.2 pixels per one diameter) for Gd CE MRA.
The Full-Width 30%-Maximum criterion is now used in the in vivo Gd CE MRA diameter measurements. MIPs of the peripheral arteries of six patients are investigated, resulting in 209 vessel diameter measurements in total MIPs and selective MIPs. In Figure 4-5, the results are plotted. In B, the absolute difference between the diameter in the selective MIP (DS) and in the total MIP (DT) (in mm) is given, versus the mean value (MEAN D) (in mm).
From Figure 4-5, it can be concluded that the diameter measurement in selective MIPs correlate very well with diameter measurements in total MIPs, with a Pearson correlation coefficient r = 0.98. A Blant Altman-analysis yielded a statistically significant mean difference of 0.15 mm between diameter measurements in selective MIPs compared to diameter measurements in total MIPs. This illustrates the fact that diameter measurements in total MIPs are slightly underestimated.
The diameter measurements in selective MIPs are now compared to diameter measurements performed in X-ray angiograms (DX), which is considered to be the standard of reference. Only diameter measurements in X-ray angiograms of the proximal station (i.e. the aorta and the iliac arteries) were possible, since a calibration was used on a catheter which was placed in the aorta. This yielded 70 measurements. Difference between the diameter measurements again are studied with a Blant Altman-analysis.
Diameter measurement in selective MIPs correlate very well with diameter measurements in the X-ray angiograms, with a Pearson correlation coefficient r = 0.92. A Blant Altman-analysis yielded a statistically significant mean difference of -0.49 mm between diameter measurements in selective MIPs compared to diameter measurements in X-ray angiograms. This implies that measurements in MRA are underestimations compared to X-ray. But an important consideration has to be noted. The standard of reference (X-ray measurement) can have a significant overestimation as well. Foreshortening of the calibration catheter and out-of-plane magnification are potential sources of error. Calibration is performed on markers placed on a catheter, which is positioned in the aorta. In case of foreshortening of this catheter of 25 degrees, the overestimation of the calibration factor will amount up to 10%.
Out-of-plane magnification occurs if the distance between the calibration object and the image intensifier is not equal to the distance between the vessel of interest and the image intensifier. The segment of the catheter with the calibration markers is usually positioned in the aorta, proximal to the iliac bifurcation. Diameter measurements are therefore not necessarily carried out in the same plane as the calibration. For a typical distance of 75 cm between the X-ray focus and the plane of interest and a typical difference of 6 cm between the plane of calibration and the plane of interest, an overestimation of 9% in calibration factor occurs. The value of these potential errors in diameter measurement in X-ray angiograms is comparable to the percent error found in diameter measurement in MR angiograms.
This study shows that automated and objective vessel diameter measurements in Gd CE MRA are possible by using the Full-Width 30%-Maximum criterion. Scan resolution currently is the limiting factor for application of this method for stenosis quantification. At least three voxels have to be present in the lumen to warrant accurate measurements, and with a scan resolution of 1.75 mm in plane voxel size, vessel diameter measurements can be performed for vessels > 5.3 mm. Introducing scan techniques with a higher scan resolution will make stenosis quantification in Gd CE MRA a reality.

Status

This project has been finished with thesis in October, 1999.

Publications

  1. Wasser MN, Westenberg J, van der Hulst VPM, van Baalen J, van Bockel JH, van Erkel AR, Pattynama PMT. Hemodynamic significance of renal artery stenosis: digital subtraction angiography versus systolically gated three-dimensional phase-contrast MR angiography. Radiology 1997; 202(2): 333-338.
  2. Westenberg JJM, van der Geest RJ, Wasser MNJM, Doornbos J, Pattynama PMT, de Roos A, Vanderschoot J, Reiber JHC. Objective stenosis quantification from post-stenotic signal loss in phase contrast magnetic resonance angiographic datasets of flow phantoms and renal arteries. Magnetic Resonance Imaging 1998; 16(3): 249-260.
  3. Westenberg JJM, van der Geest RJ, Wasser MNJM, Pattynama PMT, de Roos A, Vanderschoot J, Reiber JHC. Variations in blood flow waveforms in stenotic renal arteries by 2D phase-contrast cine MRI. Journal of Magnetic Resonance Imaging 1998; 8(3): 590-597 [erratum in Journal of Magnetic Resonance Imaging 1998, 8(6):1337]
  4. Westenberg JJM, Wasser MNJM, van der Geest RJ, Pattynama PMT, de Roos A, Vanderschoot J, Reiber JHC. Scan optimization of gadolinium contrast-enhanced three-dimensional MRA of peripheral arteries with multiple bolus injections and in-vitro validation of stenosis quantification. Magnetic Resonance Imaging 1999; 17(1): 47-57.
  5. Westenberg JJM, Wasser MNJM, van der Geest RJ, Pattynama PMT, de Roos A, Vanderschoot J, Reiber JHC. Gadolinium contrast-enhanced three-dimensional MRA of peripheral arteries with multiple bolus injections: scan optimization in-vitro and in-vivo. International Journal of Cardiac Imaging 1999; 15(2): 161-173.
  6. Westenberg JJM, van der Geest RJ, Wasser MNJM, Doornbos J, Pattynama PMT, de Roos A, Vanderschoot J, Reiber JHC. Stenosis quantification from post-stenotic signal loss in phase contrast MRA datasets of flow phantoms and renal arteries. International Journal of Cardiac Imaging 1999; 15(6): 483-493.
  7. Westenberg JJM, van der Geest RJ, Wasser MNJM, van der Linden EL, van Walsum T, van Assen HC, de Roos A, Reiber JHC. Vessel diameter measurements in gadolinium contrast-enhanced three-dimensional MRA of peripheral arteries. Magnetic Resonance Imaging 2000; 18(1): 13-22.

Abstracts

  1. Westenberg JJM, van der Geest RJ, Wasser MNJM, Doornbos J, Pattynama PMT, de Roos A, Vanderschoot J, Reiber JHC. MRA stenosis quantification in flow phantoms and renal arteries. In: Proc. ISMRM, 6th annual meeting, Sydney, 1998, volume 2, page 832, Sydney , Australia , 1998.
  2. Westenberg JJM, Wasser MNJM, van der Geest RJ, Pattynama PMT, de Roos A, Reiber JHC. Gadolinium contrast-enhanced three-dimensional MRA of peripheral arteries with multiple bolus injections: scan optimization in-vitro and in-vivo. In: Proc. ESMRMB€™98, 15th Annual Meeting, Geneva , September 17-20, 1998 abstract 451, page 182, Geneva , Switzerland , 1998.

Thesis

  • Acquisition and Analysis of Magnetic Resonance Angiographic Imaging Studies. 1999

Contact

Jos J.M. Westenberg, Ph.D.
Division of Image Processing
Department of Radiology, 1-C2S
Leiden University Medical Center
P.O. Box 9600
2300 RC Leiden
The Netherlands
Tel. +31 (0)71 526 2138
Fax. +31 (0)71 526 6801
e-mail: J.J.M.Westenberg@lumc.nl