Gradient image smoothing for metal artifact reduction (GISMAR) in computed tomography

Abstract

Metal artifacts can impair accurate diagnosis, and degrade the image quality and diagnostic value of CT-slices. In this work we propose a novel gradient image smoothing for metal artifact reduction (GISMAR) algorithm for image quality improvement in patients with hip implants, dental fillings, DBS implants and permanent seed implants. Using Image Smoothing via L0 Gradient Minimization method, a global thresholding method, and the principle of NMAR method, the authors developed a new MAR method that does not depend on access to raw projection data. To validate the authors’ approach, 2D-CT data from twenty-two patients with different metal implants were used and processed by GISMAR and three more well- known algorithms. In order to evaluate metal artifact reduction, mean CT number (HU and SD) was calculated as well as a subjective analysis with three expert observers. Image quality on images was compared using the non-parametric Friedman-ANOVA test. We conclude that GISMAR method can efficiently reduce metal artifacts using CT-slice, does not introduce new artifacts, while preserving anatomical structures.