Protocol optimization of magnetic resonance colonography for polyp detection using pig colonic phantom : influence of magnetic field strength, colonic distension technique, and MRI sequence

Abstract

Purpose: The aim of this study was to assess the diagnostic performance and image quality of magnetic resonance colonography (MRC) for colon polyp detection using pig colon phantoms and to evaluate the influence of magnetic field strength (1.5 T or 3.0 T), colonic distension technique (bright- or dark-lumen), and MRI sequence.

Materials and Methods: Six pig colon segments (60-92 cm) with 56 artificial colon polyps (0.4？1.6 cm in diameter) were placed in plastic container containing soybean oil. The colon was distended using room air for dark-lumen MRC and with tap water or a gadolinium-chelate based enema fluid for bright-lumen MRC. Each colon phantom was scanned on both 1.5 T and 3.0 T scanners using the following three sequences: axial and coronal two-dimensional (2D) fast imaging with steady-state precession (True-FISP), axial and coronal T2-weighted fat-suppressed (FS) 2D single-shot fast spin echo (SSFSE), and/or axial and coronal T1-weighted FS three-dimensional gradient-echo (3D GRE) sequences. We tried to acquire the highest spatial resolution within a 20-s acquisition time. Two radiologists evaluated the presence of polyps based on a 4-point scale and analyzed image quality with respect to artifacts, colonic wall conspicuity, polyp conspicuity, and polyp contrast using a 5-point scale. Polyp detection sensitivity and image quality were compared between image protocols or sequences using McNemar test, Friedman test, logistic generalized estimating equations, and Wilcoxon signed-rank test.

Result: For polyp detection sensitivity and image quality, MRC obtained at 1.5 T was better than that obtained at 3.0 T, and a bright-lumen technique was superior to a dark-lumen technique. Bright-lumen MRC at 1.5 T was most sensitive for polyp detection (p < 0.001) and gave the highest image quality (p < 0.05) regardless of polyp size and shape. SSFSE and 3D GRE sequences had highest sensitivity for polyp detection (83.9% and 83.0%, respectively) and image quality for bright-lumen MRC at 1.5 T.

Conclusion: The most effective sequences of MRC for polyp detection were SSFSE- or 3D GRE-based bright-lumen MRC obtained with a 1.5 T scanner. These sequences had the highest polyp detection rate and the best image quality.