Signal features of the atherosclerotic plaque at 3.0 Tesla versus 1.5 Tesla: impact on automatic classification

Abstract

Purpose: To investigate the impact of different field strengths on determining plaque composition with an automatic classifier.

Materials and methods: We applied a previously developed automatic classifier-the morphology enhanced probabilistic plaque segmentation (MEPPS) algorithm-to images from 20 subjects scanned at both 1.5 Tesla (T) and 3T. Average areas per slice of lipid-rich core, intraplaque hemorrhage, calcification, and fibrous tissue were recorded for each subject and field strength.

Results: All measurements showed close agreement at the two field strengths, with correlation coefficients of 0.91, 0.93, 0.95, and 0.93, respectively. None of these measurements showed a statistically significant difference between field strengths in the average area per slice by a paired t-test, although calcification tended to be measured larger at 3T (P = 0.09).

Conclusion: Automated classification results using an identical algorithm at 1.5T and 3T produced highly similar results, suggesting that with this acquisition protocol, 3T signal characteristics of the atherosclerotic plaque are sufficiently similar to 1.5T characteristics for MEPPS to provide equivalent performance.

Fig 1

Example images from the same subject at 1.5T (top row) and 3T (bottom row). Contrast weightings are (from left to right) T1, T2, proton density, time-of-flight, and contrast-enhanced T1.

Fig. 2

Automatic classification results for 1.5T (a) and 3T (b) images shown in Figure 1. Fibrous tissue is white, loose matrix purple, calcification blue, lipid core yellow, type I hemorrhage red and type II hemorrhage orange.

Fig. 3

Bland-Altman plots showing mean difference (solid line) and limits of agreement (+/−1.96 standard deviations; dashed lines) comparing mean areas measured with an automatic classifier at 3T versus 1.5T in 20 carotid arteries.

Fig. 3

Bland-Altman plots showing mean difference (solid line) and limits of agreement (+/−1.96 standard deviations; dashed lines) comparing mean areas measured with an automatic classifier at 3T versus 1.5T in 20 carotid arteries.

Fig. 4

Paired three-dimensional renderings (MRI-PlaqueView, VPDiagnostics, Seattle, WA) of necrotic core (yellow) and calcifications (white) with time-of-flight angiograms. In each pair, the 1.5T result appears on the left and the 3T result on the right. Correspondences in the left column were judged to be “poor” because of a large calcification only present at 1.5T (top; arrow) or offsets in the location of calcifications (middle, bottom; arrows). Correspondences for the middle column were judged to be “good” with a minor increase in the extent of the necrotic core (arrows) noted at 1.5T (top) or 3T (middle, bottom). Correspondences in the right column were judged to be “very good” with large, nearly identical necrotic cores (top; arrows), focal plaques with calcification at the bifurcation (middle; arrows), and three plaque regions with similar components (bottom; arrows).