A hybrid genetic algorithm to reduce the radiation dose in CR images. (February 2020)

• Record Type: Journal Article
• Authors: Sánchez, M.G.; Fajardo-Delgado, D.; Vidal, V.; Verdú, G.
• Is Part Of: Radiation physics and chemistry. Volume 167(2020:Feb.)
• Publication Date: 2020-02
• Subjects: Computed radiography -- Radiation dose -- Evolutionary algorithms; Radiation chemistry -- Periodicals; Radiometry -- Periodicals; Radiation -- Periodicals; Chimie sous rayonnement -- Périodiques; 539.2
• DOI: 10.1016/j.radphyschem.2019.04.025
• Languages: English
• ISSNs: 0969-806X

Abstract

Abstract: The quality of computed radiography (CR) images typically relate to patient radiation exposure. The lower the X-ray dose exposure, the higher the level of inherent noise in the CR images. In this work, we address the noise reduction problem by using an estimation of the standard deviation in CR images as an objective function to minimize. We propose a hybrid genetic algorithm for this aim, which produces improved versions of CR images. We also applied an edge-detection method based on the Canny algorithm to preserve the edges of the original CR images. We executed our proposed algorithm for CR images obtained under different radiation exposures. Experimental results show that our solution improves lower radiation CR images reaching a quality as similar to those with higher radiation doses. Highlights: We use the standard deviation in CR images as an objective function to minimize. We present a hybrid genetic algorithm which integrates the execution of two filters. The resulting images correspond to CR images obtained with a higher radiationdose.