Main Article Content

Abstract

The research presents a system which is mainly pointing to the analysis of kidney and its abnormalities like stone and cyst. The main goal of research is the classification of ultrasound (US) kidney image as normal kidney or affected one. The system with trained template is developed and user’s sample tests are verified from it. Ultrasound images contain a noise called speckle noise. It is multiplicative noise and it is introduced due to signal modification at the time of capturing an image. US images also suffers by low contrast. These issues are sorted out using filter technique and histogram equalization method. The pre-processed image is segmented using Hierarchical  k means clustering  and from it region of interest is identified. 22 features of an image are extracted and these features are trained by feed forward Artificial Neural Network (ANN) to identify the class of kidney (i.e. normal or cyst). In order to analyse the systems functionality, it is tested on a dataset of ultrasound images of two classes. The analysis performance is based on two parameters first is accuracy and second is precision, which results in 87.5% and 100% respectively.

Article Details

How to Cite
R.Janarthanan, G.Kalaiyarasi, M.MuthuSowmya, M.NoorAshma, & S.Marlarkhodi. (2021). Design and analysis performance of kidney stone and cyst detection from ultrasound images. International Journal of Intellectual Advancements and Research in Engineering Computations, 7(1), 18–27. Retrieved from https://ijiarec.com/ijiarec/article/view/158