Integrating Bilateral Filtering with Anisotropic Diffusion – A model for Enhancing Diagnostic Quality of Multi-Noise affected COVID-19 Lung Ultrasound Images

Abstract

Lung Ultrasound (LUS) has become a widely accepted technique for Covid-19 prognosis and diagnosis. This is due to the non-risk factors of ultrasound (US) and the emergence of portable Point of Care equipments. However, multiple types of noise such as gaussian, impulse and speckle noises usually affect ultrasound images.  Distinct filters are required for handling each type of noise. Developing a single filter capable of eliminating nearly all types of noise would be highly significant. Considering the characteristics of different filters used for noise reduction, a composite filter has been developed to address all types of noise. The integration of bilateral filters in estimating the conduction coefficient of Anisotropic Diffusion (AD) filters facilitates this accomplishment. To eliminate impulse noise, median filters are incorporated as an intermediate process. The filter performance is quantitatively verified across various types of noise, including mixed Gaussian impulse noise and speckle noise, using performance metrics such as Peak Signal-to-Noise Ratio (PSNR), Mean Squared Error (MSE) and Structural Similarity Index (SSIM). Compared to the existing method of employing individual filters for noise removal, the suggested combined filter shows superior performance in qualitative and quantitative analysis. Resultant images exhibit higher Peak Signal-to-Noise Ratio (PSNR), minimized Mean Squared Error (MSE) and yield improved outcomes.  The performance of the proposed filter is evaluated using images affected with speckle noise and resulted in highly improved results. Amidst the rapidly evolving field of artificial intelligence and its application in healthcare, more accurate predictions can be made using these cleaner images, thus improving the performance and reliability of AI algorithms.