Measurement of Plaque Burden using 3D SNAP vessel wall MRI

Niranjan Balu¹, Jie Sun¹, Daniel S Hippe¹, Huijun Chen², and Chun Yuan¹

¹Radiology, University of Washington, Seattle, WA, United States, ²Tsinghua University, Beijing, China, People's Republic of

Synopsis

We have developed a single sequence (SNAP) for identification of high-risk features such as stenosis, intraplaque hemorrhage and juxtaluminal calcification. However measurement of plaque burden, an early marker for disease, required a separate black-blood scan. In this work, we demonstrate that plaque burden can be measured using an alternate reconstruction (SNAP2). In patients with atherosclerosis, we compare vessel wall area measurements on SNAP2 to established black-blood MRI measurements and show that our method provides results comparable to established plaque MRI methods.

Introduction

Atherosclerotic plaque features such as intraplaque hemorrhage (IPH) and juxtaluminal calcification (JCA) are considered to indicate high-risk status [1]. Carotid plaques with these features are more likely to cause stroke due to thromboembolism. Recently a phase-sensitive inversion-recovery (IR) based sequence (SNAP) [2] was extended to include a PD weighting (SNAP2) that can identify calcification and show its JCA status [3]. Even though IPH and JCA can be identified in this manner, measurement of plaque burden which is an earlier marker of disease than IPH/JCA, required a separate black-blood sequence. In this work, we show that plaque burden can be reliably measured using the SNAP2 formulation alone.

Aims

To compare plaque burden measurement between SNAP2 against 3D black-blood (3D BB) carotid plaque MRI.

Materials and methods

Image acquisition: Bilateral arteries (N=14) from 7 patients with 50-79% stenosis by ultrasound were examined with MRI. Using an 8 channel phased array carotid coil, SNAP MRI was acquired with parameters: TR/TE 10/4ms, flip angle 11/5 (corresponding to ⍺ and θ in fig 1), TI 500ms, Resolution 0.8x0.8x0.8mm, FOV (coronal) 16x16x3.2cm, turbo factor 98, scan time 5 min. Thus the first image (I1) was T1-weighted and the second image (I2) was PD weighted. Image reconstruction: A polarity function P(x,y) which takes values (-1 or +1) depending upon the longitudinal magnetization was calculated using I1(x,y) and I2(x,y) as $$P(x,y)= (I_1 (x,y)I_2^* (x,y))/‖I_1 ‖‖I_2 ‖$$ where * represents complex conjugation. T1-weighted corrected real image was then obtained as $$$S_1 (x,y)=P(x,y)‖I_1 (x,y)‖$$$ and PD-weighted corrected real image was obtained as $$$S_2 (x,y)=P(x,y)‖I_2 (x,y)‖$$$. A sigmoid function was used in reconstructing S2(x,y) such that gray-blood PD-weighted vessel wall image (SNAP2) was available for plaque burden measurements. A previously validated 3D black-blood sequence [4] was used for reference standard plaque burden measurements. Imaging parameters were: TR/TE 10/4ms, flip angle 6, Resolution 0.8x0.8x0.8mm, FOV (coronal) 25x16x3.2cm, turbofactor 30, scan time 3.5 min. Patient image review: The coronally acquired SNAP2 and 3D BB were reformatted into axially oriented 2mm slices. Lumen and outer wall boundaries were drawn separately on SNAP2 and 3D black-blood MRI by two different reviewers. Slices were matched across the two sequences based on their location relative to the carotid bifurcation. Percent wall volume (%WV) was computed as wall area/total vessel area x 100% and summed over slices per artery. Wall, lumen and total vessel areas were also averaged for each artery. Artery level measurements were compared between the two sequences. Statistics: The non-parametric bootstrap and percentile method was used to calculate 95% confidence intervals (CIs) for the ICCs. Overall bias was tested using a permutation test.

Results

Common coverage reviewed ranged from 4 to 30 slices (median: 19, mean: 17.4 ± 10.2 slices). On an artery level, there was no significant difference in wall area between SNAP2 and 3D BB measurements (32.6 ± 8.3 mm2 vs 35.8 ± 8.1 mm2, p=0.17). Figure 2 shows representative slices of SNAP2 plaque boundaries vs 3D BB boundaries. Figure 3 shows corresponding Bland-Altman plot. There was good ICC for lumen and total vessel areas but ICC was low for wall area. However %WV measure of plaque burden showed improved ICC (table 1).

Discussion

Measures of plaque burden such as wall area and %WV were comparable between SNAP2 and 3D BB. Lumen area and total vessel area tended to be larger on SNAP2 which could be due to reviewer variability or difference in sequence contrast and blood suppression capabilities.

Conclusions

Plaque burden can be measured using SNAP2 and was comparable to plaque measurements on 3D-BB. Thus use of SNAP2 along with other SNAP derived image weightings (SNAP2 and I2) may provide a time-efficient means for comprehensive high-risk plaque detection using a single sequence.

Acknowledgements

No acknowledgement found.

References

[1] Saam Radiology 2007; 244(1):64-77, [2] Wang MRM 2013; 69(2):337-45. [3] Balu ISMRM 2015. [4] Balu MRM 2011; 65(3):627-37

Figures

Figure 1: Sequence diagram showing two images acquired in a single acquisition, a highly T1-weighted I1(x,y) and the PD-weighted I2(x,y).

Figure 2: Plaque boundaries drawn on SNAP2 compared to plaque boundaries on 3D-BB. Red-lumen, blue-outerwall, * denotes calcification. Other SNAP derived images, SNAP1 and I2 are also shown.

Figure 3: Bland-Altman plot comparing mean wall area between SNAP2 and the standard (3D-BB). The dashed lines indicate the mean differences and the dotted lines indicate the limits of agreement.

Table 1: Agreement of morphology measurements.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)

2547