Reduced Image Noise at Low-Dose Multidetector CT of the Abdomen with Prior Image Constrained Compressed Sensing Algorithm

Abstract

To assess the effect of prior image constrained compressed sensing (PICCS) on noise reduction and image quality at low-dose computed tomography (CT). This HIPAA-compliant institutional review board-approved retrospective study was performed by using DICOM CT colonography data sets obtained in 20 adult patients. Informed consent was waived. Low-dose CT colonography was performed with 64-detector CT by using the standard protocol with mean effective dose per series of 3.06 mSv (range, 1.4-7.7 mSv). PICCS was applied to standard filtered back-projection (FBP) series. For FBP and PICCS series, mean and standard deviation (SD) of attenuation were obtained with 100-mm(2) circular region of interest (ROI) at six sites (240 soft-tissue, colonic gas, and subcutaneous fat measurements). Two abdominal radiologists reviewed two- and three-dimensional CT colonography displays and graded image quality with a five-point scale. Phantom studies were performed to compare spatial resolution and image quality between FBP and PICCS. Mean image noise and image quality scores were calculated and compared for clinical and phantom data sets. Bland-Altman, generalized estimating equation regression model, and Student t tests were used to obtain limits of agreement and to compare noise ratios and subjective image quality. Mean SD of attenuation (image noise) for ROIs was 38.0 for FBP and 12.2 for PICCS, corresponding to a noise-reduction factor of 3.1 (P < .001). Average noise reduction was 3.3 for soft tissue, 2.8 for air, and 3.0 for fat attenuation. Attenuation did not substantially change between FBP and PICCS images. Average two-dimensional image quality was 2.45 for FBP and 3.4 for PICCS (P < .001). Average three-dimensional image quality at three sites in the colon was 3.5 for FBP and 3.7 for PICCS (P = .34). Phantom data sets revealed no loss of spatial resolution in a line phantom and reduced noise in a liver tumor phantom when PICCS was compared with FBP. Application of PICCS to standard FBP low-dose multidetector CT abdominal images results in substantial noise reduction and improved image quality.