Feasibility of ultra-low-dose multi-detector-row CT-colonography:: Detection of artificial endoluminal lesions in an in-vitro-model with optimization of image quality using a noise reduction filter algorithm

Purpose: To assess the most favorable slice thickness in Multi-Detector-Row CT-colonography (MDCTC), and the feasibility of dose reduction in an in-vitro-setting as well as the possibility of optimization of image quality using a noise reduction filter algorithm. Materials and Methods: 18 artificial lesions with sizes from I to 8 mm were randomly positioned in two cleansed pig colons. At a "Somatom Plus 4 Volume Zoom", six scanning protocols using a slice collimation of 2.5, 1, and 1 mm with a reconstructed slice thickness of 3, 3, and 1.25 min were performed with tube currents of 100, and 10 mAs, respectively. Using a non-commercial software, a non-linear Gaussian filter was used to minimize image noise. Image noise was assessed before and after application of the filtering process. Using a threshold of -750 HU, two blinded readers analyzed the virtual colonography in respect to lesion location, size, and shape. Artifacts were noted. An automated detection system was evaluated. Results: Using 10 mAs, a ten-fold dose reduction was achieved. After application of the mathematical filter, image noise was reduced by 45-80% for 100 mAs, and by 50-70% for 10 mAs scans. Only with a slice thickness of 1.25 mm, all lesions could be detected. The definition of lesion size and shape was more accurate with higher mAs. Only minor noise artifacts were noted on low-dose images. The automated polyp detector marked not more than 60% of artificial lesions. Conclusion: MDCTC benefits from narrow slice collimation. In an in-vitro-model, a significant dose reduction is achievable with preservation of a high lesion detection rate. The noise reduction filter algorithm improved image quality substantially.