A Computerized Diagnostic System For Kidney Region Segmentation Aimed At Identifying Abnormalities In Ultrasound Pictures of The Kidney

Acquiring medical pictures via computational methods has been widely utilised to better identify and treat disease. In many cases, it is impossible for doctors to make a conclusive decision about treatment without first obtaining such photographs. A small subset of imaging methods provides stable groundwork for image acquisition. One such modality, ultrasound, is well-liked for its low cost, lack of invasiveness, and lack of radiation risk. This modality's low cost and high efficacy have made it widely accessible in India. The study's overarching goal is to investigate the potential for developing a computer-aided diagnosis (CAD) system to aid radiologists in the detection of kidney abnormalities in ultrasound pictures. Images, regions, and features all need to be pre-processed, segmented, and extracted and classified in order to make CAD a reality. In the module responsible for preparing images, a method for effectively and robustly reducing speckle is proposed. Speckle is a type of noise that decreases image quality and is unique to ultrasound images because of the way in which ultrasound images are obtained. The suggested approach is implemented in the curvelet domain and utilizes Maximum Aposteriori Probability (MAP) estimation of coefficients under the a priori probability assumption of a Gaussian distribution in conjunction with diffusion filtering through the Perona Malik Anisotropic Diffusion filter (PMAD). The proposed method has been evaluated using approximated quantitative metrics as peak signal-to-noise ratio, structural similarity, and coefficient of correlation. The effectiveness and robustness of the suggested work in terms of quantitative metrics is further confirmed by a comparison with the recent studies.