Yinsheng Li MD, Presenter: Nothing to Disclose

Perry J. Pickhardt MD, Abstract Co-Author: Co-founder, VirtuoCTC, LLC Stockholder, Cellectar Biosciences, Inc

Meghan G. Lubner MD, Abstract Co-Author: Nothing to Disclose

Guang-Hong Chen PhD, Abstract Co-Author: Research funded, General Electric Company Research funded, Siemens AG Research funded, Varian Medical Systems, Inc Research funded, Hologic, Inc

Metal artifacts often contaminate CT images and hinder medical diagnosis around the metal implant and surrounding soft tissue. The purpose of this study was to assess a novel compressed sensing based method developed to remove metal artifacts from clinical CT images.

The compressed sensing metal artifact reduction (CS-MAR) algorithm was implemented and retrospectively applied to DICOM image data sets from 40 human subjects. Metal artifact levels were qualitatively evaluated based on the perceived metal artifact level and quantitatively evaluated by measuring the standard deviation of CT numbers in soft tissue surrounding the metal implants. The qualitative and quantitative results were compared between both the original clinical images and processed images.

Qualitative observation demonstrated that, for all 40 subjects in this study, the metal implants appeared properly reconstructed after the CS-MAR algorithm was applied. The shading and streaking artifacts surrounding the metal implants were significantly reduced to enable clear visualization of the surrounding soft tissue. Quantitatively, the standard deviation of the CT numbers in the surrounding soft tissue regions were significantly reduced due to the mitigation of both metal streaks and shading artifacts. Quantitative measurements for two subjects are presented as examples. In the first example , the standard deviation of the CT numbers for three regions of interest (ROIs) proximal to the metallic implant were reduced from 61, 56, and 63 HU to 36, 27, and 31 HU respectively. The standard deviation of the CT numbers for the three ROIs farther away from metallic implants were reduced from 56, 53, and 64 HU to 37, 32, and 39 HU respectively. For the second sample case, the standard deviation of CT numbers was reduced from 124, 111, and 137 HU to 47, 27, and 49 HU respectively for the three ROIs proximal to the metal implants, while the standard deviations of CT numbers were reduced from 102, 57, and 86 HU to 31, 30, and 28 HU respectively for the three ROIs further away from the metal implants. 

The CS-MAR algorithm can be applied to any clinical CT dataset to reduce metal artifacts, enabling clear visualization of both metal implants and surrounding soft tissues.

The CS-MAR algorithm can be applied to any clinical CT cases with metal artifacts directly from DICOM images to improve diagnostic performance. 

Li, Y, Pickhardt, P, Lubner, M, Chen, G, Compressed Sensing Based Metal Artifact Reduction (CS-MAR) Algorithm.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14015814.html