The aim of this study was to assess changes in acquisition time, image quality and evaluation of pancreatic cysts when applying CS to a 3D MRCP sequence. Thirty subjects (17F; 13M) undergoing MRCP for evaluation of pancreatic cyst(s) were prospectively recruited and underwent 3D MRCP and CS 3D MRCP (CS factor = 2) on a 3T scanner. The acquisition time was recorded. Two experienced radiologists independently recorded quality of the images, presence of artifacts, visualization of the main pancreatic duct, bile ducts and index pancreatic cyst using a five-point scale. Presence of mural nodules and septations in the cyst, size of the cyst and caliber of the main pancreatic duct were also recorded. A paired sample t-test was used to compare the acquisition time of 3D MRCP and CS 3D MRCP. Image quality metrics and visualization of cyst features were compared with Wilcoxon signed-rank test and McNemar test. The mean acquisition time of CS-3D-MRCP (150 ± 63 s) was significantly lower than that of 3D-MRCP (317 ± 104 s; P < 0.001). The median score of overall quality (reader 1, 3.7 ± 1.0 vs. 3.4 ± 1.1, P = 0.11; reader 2, 3.8 ± 1.0 vs. 3.7 ± 1.1, P = 0.36), artifacts and visualization of the bile ducts were not significantly different between 3D-MRCP and CS-3D-MRCP. There was no significant difference in the visualization score of the index pancreatic cyst (reader 1, 4.2 ± 0.9 vs. 4.1 ± 0.9, P = 0.42; reader 2, 4.2 ± 0.4 vs. 4.0 ± 0.7, P = 0.27) and no difference in the assessment of cyst features. Applying CS to 3D-MRCP yields a two-fold reduction in acquisition time with comparable image quality and visualization of key pancreatic cyst features.
Keywords: Compressed sensing; MRCP; Pancreatic cyst.
Copyright © 2018 Elsevier Inc. All rights reserved.