Comparison of Denoising Methods Applied to CTA Images of 3D Segmentation of Aortic Dissection
Maya Fitria, Ramzi Adriman, Cosmin Adrian Morariu, Josef Pauli · 2019
Noise is a disturbance of the image quality that characterized by the presence of pixels that are not desired in the image. Removing or reducing the appearance of the noise is still a challenging problem in the image processing field. We present several denoising methods in this work, such as anisotropic filter, bilateral filter, and Yaroslavsky filter. The objective of this work is to reduce the distortion in the medical images, particularly in Computed Tomography Angiography (CTA) images. It is essential to take care of valuable information about human organ (e.g., in the aorta) carried by medical images since they will be used to diagnose. This work is carried out in the aortic segmentation system which can detect and display the dissection part of the aorta. We embedded the denoising methods in the system in order to enhance the detection of the aortic contour, which subsequently leads to more accurate aortic segmentation results. We compare the performance of each method by using the aortic segmentation result represented in the form of the Dice Similarity Coefficient, and also the visual quality image. The results showed that the segmentation results of all denoising techniques presented in this work are encouraging, although not all the visualization of processed images are well maintained. Anisotropic diffusion is one noise reduction technique that obtained the best in both segmentation result and denoised image.