Noise reduction in CT perfusion images: A bilateral filter trial

Computed Tomography (CT) perfusion images are acquired in short time period and in order to lower patient radiation dose. Low x-ray photon flux and energy are preferred in such CT exposures. Obtainedimages contain more noise than classical CT images which are usually obtained via using high photon flux and high energy. Perfusion images are used for obtaining functional perfusion maps. Singular value decomposition (SVD) based deconvolution algorithms are mainly used for this purpose. SVD based algorithms are very sensitive to noise because of their intrinsic nature. To denoise perfusion images with noise reduction filters is a very common application in order to obtain functional maps with better image quality to help physician through diagnostic process. In this study, regarding to itssuperior noise reduction and detail preservation capabilities, bilateral filter is chosen for noise reduction. Bilateral filter parameters are optimized via an emprical method on CT perfusion images of suspected acute stroke patiens. Results are evaluated and compared by visual means, peak signal noise ratio, mean square error metrics and region of interes analyses on obtained perfusion maps. It is observed that high magnitude of the present noise is denoised and details containing contrast bolus are preserved.