Application of compressed sensing to 3D magnetic resonance cholangiopancreatography for the evaluation of pancreatic cystic lesions

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.