Francois Pontana MD, Presenter: Nothing to Disclose

Julien Pagniez, Abstract Co-Author: Nothing to Disclose

Thomas G. Flohr PhD, Abstract Co-Author: Employee, Siemens AG

Jean-Baptiste Faivre MD, Abstract Co-Author: Nothing to Disclose

Jacques Remy MD, Abstract Co-Author: Research Consultant, Siemens AG

Martine J. Remy-Jardin MD, PhD, Abstract Co-Author: Research grant, Siemens AG

To assess noise reduction and subjective image quality on chest computed tomography (CT) examinations reconstructed with an iterative reconstruction algorithm.

32 consecutive chest CT angiograms were reconstructed both with regular filtered back projection (FBP) and an iterative reconstruction technique (IRIS; Siemens) with 3 and 5 iterations. IRIS uses images reconstructed with a dedicated convolution kernel maintaining high spatial frequencies at the expense of increased image noise as input for an iterative loop in the image domain. Regularization of the update images by constraining and correlating neighbouring image signals controls image noise and sharpness almost independently. A novel regularization prior which is a linear combination of a smoothing and an edge enhancing term allows one to establish contrast dependent sharpness by nonlinear and anisotropic filtration. In this way, edges are preserved or even enhanced, while the image noise in more homogeneous areas is significantly reduced. Moreover, the familiar noise texture of typical CT convolution kernels is reproduced within the regularization procedure.

Objective image noise was reduced with 3 iterations (trachea: 13.98; descending aorta: 32.87 HU) and 5 iterations (trachea: 10.75; descending aorta: 26.08 HU) compared with FBP (22.26 and 48.0 HU; p<0.0001), corresponding to a mean noise reduction of 37.2% (trachea) and 31.5 % (aorta) with 3 iterations and 51.7% (trachea) and 45.7% (aorta) with 5 iterations. With both 3 and 5 iterations, subjective image noise was lower than with FBP (p<0.005), without loss of diagnostic information as assessed by the comparative analysis of individual CT features of lung infiltration (i.e., ground-glass opacities, ill-defined micronodules, nodules, septal and nonseptal lines and areas of decreased lung attenuation) depicted at preselected levels-of-interest on the 3 series of reconstructions (p=1).

Compared with regular FBP, iterative reconstructions enable significant reduction of image noise without loss of diagnostic information, thus having the potential to decrease radiation dose during chest CT examinations.

Assessing the range of noise reduction without loss of diagnostic information is a mandatory step before introducing iterative reconstructions for radiation dose reduction.

Pontana, F, Pagniez, J, Flohr, T, Faivre, J, Remy, J, Remy-Jardin, M, Noise Reduction on Chest CT Examinations: Impact of Iterative Reconstructions.  Radiological Society of North America 2010 Scientific Assembly and Annual Meeting, November 28 - December 3, 2010 ,Chicago IL. http://archive.rsna.org/2010/9004492.html