Segmentation of carotid plaque using multicontrast 3D gradient echo MRI

Purpose: To evaluate the performance of automatic segmentation of atherosclerotic plaque components using solely multicontrast 3D gradient echo (GRE) magnetic resonance imaging (MRI). Materials and Methods: A total of 15 patients with a history of recent transient ischemic attacks or stroke underwent carotid vessel wall imaging bilaterally with a combination of 2D turbo spin echo (TSE) sequences and 3D GRE sequences. The TSE sequences included T1-weighted, T2-weighted, and contrast-enhanced T1weighted scans. The 3D GRE sequences included time-offlight (TOF), magnetization-prepared rapid gradient echo (MP-RAGE), and motion-sensitized driven equilibrium prepared rapid gradient echo (MERGE) scans. From these images, the previously developed morphology-enhanced probabilistic plaque segmentation (MEPPS) algorithm was retrained based solely on the 3D GRE sequences to segment necrotic core (NC), calcification (CA), and loose matrix (LM). Segmentation performance was assessed using a leave-one-out cross-validation approach via comparing the new 3D-MEPPS algorithm to the original MEPPS algorithm that was based on the traditional multicontrast protocol including 2D TSE and TOF sequences. Results: Twenty arteries of 15 subjects were found to exhibit significant plaques within the coverage of all imaging sequences. For these arteries, between new and original MEPPS algorithms, the areas per slice exhibited correlation coefficients of 0.86 for NC, 0.99 for CA, and 0.80 for LM; no significant area bias was observed. Conclusion: The combination of 3D imaging sequences (TOF, MP-RAGE, and MERGE) can provide sufficient contrast to distinguish NC, CA, and LM. Automatic segmentation using 3D sequences and traditional multicontrast protocol produced highly similar results.