Performance evaluation of improved median-modified Wiener filter with segmentation method to improve resolution in computed tomographic images

Abstract

This study aimed to model and apply an improved median-modified Wiener filter (MMWF) with region growing (RG) segmentation technique in computed tomography (CT) images to improve the noise and blurring effects. To model the improved MMWF, the MATLAB program was used to provide an additional application of median filter and RG techniques from the smoothening process of Wiener filter. In addition, the kernel size of the improved MMWF was set to 7 x 7 and applied to abdominal CT images of the acquired whole body phantom PBU-50 (Kyoto Kagaku, Japan) with tube currents of 50, 100, 200, and 300 mAs, and quantitatively compared to images with conventional MMWF and without MMWF (noisy). Compared with noisy images, the coefficient of variation (COV) and contrast-to-noise ratio (CNR) of the conventional and improved MMWFs improved by approximately 3.41 and 3.32 times, respectively, for the images at all tube current conditions. The improved MMWF showed improved separation between different tissues compared with noisy images even though the smoothening was performed on images with low-dose conditions (50 and 100 mAs). Moreover, a comparative evaluation with other conventional filters (median, Wiener, total variation, and non-local means) demonstrated the improved MMWF performance at low-dose conditions. The COV and CNR of the improved MMWF were overall balanced; however, the improved MMWF performed better than other filters to restore the boundary signal between the two tissues. In conclusion, the improved MMWF with RG technique could acquire CT images with improved characteristics by removing the noise and blurring effects.