INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 9, ISSUE 09, SEPTEMBER 2020 ISSN 2277-8616 231 IJSTR©2020 www.ijstr.org Effectiveness Of A Selected Mean Filter Algorithm To Reduce Noise In Fluoroscopy Images Jhon Hadearon Saragih, Choirul Anam, Wahyu Setia Budi, Ummu Mar'atu Zahro, Geoff Dougherty Abstract : This study aims to implement noise reduction algorithm with a selected mean filter (SMF) and to investigate its computation time in the denoising process on X-ray fluoroscopy images. The SMF was the mean filter (MF) technique, but in its application, selected pixels within threshold value were only used to calculate the average pixel value. The effectiveness of SMF was then compared to well-known filters, such as adaptive mean filter (AMF) and bilateral filter (BF). The notebook of Acer Nitro 5 Intel Core i5-8300H 2.3 GHz with 8GB RAM, Graphic Processor Unit (GPU) Nvidia Geforce GTX 1050 4GB, and the Windows 10 Home operating system with SSD M.2 NVMe 2280 256GB were utilized. The algorithm was implemented using Matlab R2019b. The fluoroscopy images of NEMA SCA & I Cardiovascular Fluoroscopic Benchmark Phantom with size of 512 x 512 pixels were filtered, exposure factors of 69.92 kV and 583 mA and a dose-area product (DAP) of 1,660 mGy-cm 2 with a field of view (FOV) of 25 cm. In addition, image quality of the filtered images was assessed, including noise level, signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and spatial resolution. The results showed that by using the SMF, the higher improvement of image quality in terms of noise level, SNR, CNR, and spatial resolution compared to AMF and BF, was achieved. The time needed by SMF to process an image was about 0.36 seconds, while the AMF and BF are 10.6 and 1.4 seconds, respectively. The SMF was as fast as a traditional MF, which only need 0.33 seconds for an image. Index Terms : Selected Mean Filter, Noise Reduction, Fluoroscopy, Cardiovascular, Image Quality, Radiation ——————————  —————————— 1. INTRODUCTION FLuoroscopy plays an important role in medical imaging, and has been widely used in the clinical examination of patients and in intervention procedures (angiography). Fluoroscopy employs ionizing radiation so that it poses a potential danger to patients [1], despite its benefits as a gold standard for diagnosis in many cases. Exposure to ionizing radiation must always be kept as low as reasonably achievable (ALARA), by providing the minimum amount of radiation required in order to provide useful diagnostic results [2]. Reducing radiation exposure is an essential challenge related to fluoroscopy X-ray imaging. However, when the radiation dose is minimized, the resulting high noise results in a reduced image quality that leads to a decreased accuracy of clinical diagnosis. Thus, some form of image denoising is required. A noise reduction algorithm is considered as an effective technique in reducing the dose received by patients [3], [4]. Previously, a study on non-linear filters showed that the bilateral filter (BF) could denoise images without causing a deterioration in the spatial resolution of the images [5]. It was observed that the BF produced images of similar quality to those from full-dose exposure, but with half the dose exposure. The average time required to apply the BF to 512 x 512 pixels image using a portable computer (Intel processor Core-2-due 2.6 GHz, RAM 4 GB, Cash Memory 3 MB) and a MATLAB (version 7.12 R2011) implementation was about 25 seconds [5]. Similar studies on the effectiveness of BF algorithm and other algorithms, such as adaptive mean filter (AMF), have also been carried [6]–[10]. Available denoising algorithms require a heavy computational burden and a long processing time, so that it is difficult to implement them in clinical X-ray fluoroscopy which typically results in 15-30 frames per second. This may be mitigated by using high-speed computer technology or a more efficient denoising algorithm. A selective mean filter (SMF) has been proposed as a fast filter which reduces noise while maintaining the spatial resolution of an image [11]. However, the algorithm has only been implemented in computed tomography (CT) images. The aim of this study is to implement SMF for X-ray fluoroscopy images, and to investigate the resulting computational time and image quality. 2 RESEARCH METHODS 2.1 SMF algorithm The SMF was recently introduced and is based on the simple mean filter technique. However, not all neighboring pixels are used to calculate the average pixel value, rather it is done selectively based on a threshold value (h). That is, if the differences of the values of neighboring pixels in a kernel and the central pixel are greater or smaller by a threshold value h, then the neighboring pixel is not included in the selective mean. The selection of pixels uses equation (1). (1) Noise reduction in the image is computed using the equation (2), (2) ————————————————  Postgraduate Program, Department of Physics, Faculty of Sciences and Mathematics, Diponegoro University, Jl. Prof. Soedarto SH, Tembalang, Semarang 50275, Central Java, Indonesia. E-mail: jhonhsaragih.2019@fisika.fsm.undip.ac.id  Department of Physics, Faculty of Sciences and Mathematics, Diponegoro University, Jl. Prof. Soedarto SH, Tembalang, Semarang 50275, Central Java, Indonesia. E-mail: anam@fisika.fsm.undip.ac.id