VLSI Implementation and Analysis of Kidney Stone Detection from Ultrasound Image by Level Set Segmentation and MLP-BP ANN Classification

Abstract

In recent years, there is an increase in the count of individuals suffering from kidney abnormalities. Kidney stone prevalence has increased both in men and women, across all age groups, racial/ethnic groups. According to the recent statisticscal report, the vulnerability of kidney stone abnormality even surpasses the effects of several chronic diseases, including diabetes, coronary heart disease, and stroke. This inflicts a need for early detection and accurate diagnosis of kidney stones. Urologists undergo enormous stress at the time of surgery related to stone removal in order to precisely locate the stones, which may be scattered. Kidney abnormality may also indicate the formation of stones, cysts, cancerous cells, and blockage of urine, etc. Currently available scanning approaches in hospitals such as Ultrasound (US) imaging, MRI, and CT scanners, do not help in easy and quick diagnosis of the minute stones in the initial stage, as well as multiple stones present in the scanned images due to low contrast and speckle noise. Thus, to remove speckle noise in ultrasound images preprocessing is applied. Reaction and diffusion (RD) level set segmentation is applied two times, first to the segment kidney portion and second to segment the stone portion. The extracted region of the kidney stone after segmentation is applied with Symlets, Biorthogonal, and Daubechies lifting scheme wavelet subbands with higher vanishing moments to extract energy levels. These energy levels give an indication about the presence of stone, which significantly vary from that of normal energy level. These energy levels are trained by multilayer perceptron (MLP) and back propagation (BP) ANN to identify the type of stone with an accuracy of 97.8 % and real time implementation is done using Verilog on Vertex-2Pro FPGA.