Roni Shreter MD, Presenter: Consultant, Medic Vision Ltd

Reuven Shreiber MD, Abstract Co-Author: Employee, Algotec Limited

Doron Fischer BSc, MD, Abstract Co-Author: Nothing to Disclose

Ahuva Engel MD, Abstract Co-Author: Nothing to Disclose

It has been postulated that image reconstruction algorithms that increase the Signal-to-Noise Ratio (SNR) of low-dose CT images will reduce Low Contrast Detectability (LCD). This research evaluates a novel image reconstruction algorithm, showing improved LCD in low-dose SNR-enhanced images.

Two CT performance phantoms (the American College of Radiology (ACR) CT accreditation and Catphan 600), containing 2.0-15.0 mm (diameter) low-contrast targets at 0.3-1.0% contrast levels, were scanned on the 64-slice GE Discovery 690 and Philips Brilliance 64 CT scanners, respectively. 32 thin-slice datasets were acquired in axial or helical scanning modes at dose levels of 8-50 mGy, with parameter combinations simulating typical acquisitions. An additional 32 corresponding datasets were generated by applying 3D non-linear post-processing (SafeCT from Medic Vision Imaging Solutions) to the original datasets, using default abdomen processing parameters. All datasets were reformatted to slice thicknesses of 0.625-5.0 mm. 2 slices were extracted from each dataset for review. The 128 images were presented in random order to 5 experienced CT readers (3 radiologists and 2 physicists) who were requested to specify the smallest detectable low contrast targets.

All processed images showed a decrease of at least 70% in image noise relative to their corresponding original images. The related-samples Wilcoxon signed rank test was performed, showing that LCD was significantly higher in the processed images (p <0.01 for the ACR and p <0.0001 for the Catphan phantoms) than in the original images. Highest LCD improvements were observed at low-dose levels (200%, p <0.0001 for 100 mAs) and at low-contrast levels (150%, p <0.0001 for 0.3% contrast level). In some cases, processed images showed higher LCD compared to the original images acquired at double the dose.

3D non-linear post-processing may improve not only the SNR of the CT image, but also the LCD. This was confirmed over a wide range of CT parameters and conditions, suggesting significant potential for CT dose reduction.

Image quality and radiation dose are of prime concern in CT imaging today. This study demonstrates improved quality of low-dose images in terms of SNR and LCD using a novel post-processing algorithm.

Shreter, R, Shreiber, R, Fischer, D, Engel, A, Enhancement of Signal-to-Noise Ratio and Low Contrast Detectability in Low-Dose CT Images Using Nonlinear Postprocessing.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11004986.html