Hye Sun Na, Presenter: Employee, General Electric Company

Akira Hagiwara, Abstract Co-Author: Employee, General Electric Company

Brian Edward Nett PhD, Abstract Co-Author: Employee, General Electric Company

To assess windmill artifact mitigation performance for a new reconstruction algorithm at large cone angles

Windmill artifacts are caused by insufficient sampling in the z-direction and objects with high z-gradient. In this work, a new reconstruction algorithm is evaluated for correction of windmill artifacts which occur at large cone angles in axial scans. A phantom was simulated with wires at a slope of 4:1, axial to transaxial, to assess the severity of windmill artifacts and to measure z-axis spatial resolution before and after the correction is applied. The images were reconstructed with a pixel size of 0.375mm.

A windmill artifact metric was developed, which is the sum of the image after applying a mask and a Sobel edge filter. The mask was created by applying two dilations from the ground truth image. A 20x20 pixel matrix was used as the structuring element for the first dilation to mask out the wire and a 140x140 pixel matrix was used for the second dilation to mask out regions beyond where the artifacts are generated around the wire. In this manner, images with severe windmill artifacts will receive a higher value for this metric. In the FDK images the level of artifact increases as the cone angle increases; the ratio of the mean artifact metric values for central (≤ ±1.5°) to outer cone angles (±3.1-6°) is 0.23. In the improved recon the artifact level in the outer cone angles is lower than the artifact level in the central cone angles of the FDK images. The ratio of the mean artifact metric for outer cone angles [improved/FDK] is 0.18. Additionally, the improved recon reduces artifacts in the central cone angle case. The ratio of the mean artifact metric [improved/FDK] is 0.15. The z-resolution was measured at the same locations. With the new recon the mean MTF50% and MTF10% is 100% and 99% of the FDK recon and varies from 96.2%-105.0% and 97.9%-102.7%, respectively.

At large cone angles, a new reconstruction algorithm reduces windmill artifacts without significantly degrading z-resolution.

A new reconstruction algorithm, which reduces windmill artifacts in clinical head CT scans at large cone angles, is an enabling technology for single axial scans which cover all neuro anatomy.

Na, H, Hagiwara, A, Nett, B, Mitigation of Windmill Artifacts at Large Cone Angles.  Radiological Society of North America 2013 Scientific Assembly and Annual Meeting, December 1 - December 6, 2013 ,Chicago IL. http://archive.rsna.org/2013/13016857.html