Comparison of multilayer perceptron training algorithms for portal venous doppler signals in the cirrhosis disease

Abstract

In this study, we developed an expert diagnostic system for the interpretation of the portal vein Doppler signals belong the patients with cirrhosis and healthy subjects using signal processing and Artificial Neural Network (ANN) methods. Power spectral densities (PSD) of these signals were obtained to input of ANN using Short Time Fourier Transform (STFT) method. The four layered Multilayer Perceptron (MLP) training algorithms that we have built had given very promising results in classifying the healthy and cirrhosis. For prediction purposes, it has been presented that Levenberg Marquardt training algorithm of MLP network employing backpropagation works reasonably well. The diagnosis performance of the study shows the advantages of this system: It is rapid, easy to operate, noninvasive and not expensive. This system is of the better clinical application over others, especially for earlier survey of population. The stated results show that the proposed method can make an effective interpretation and point out the ability of design of a new intelligent assistance diagnosis system.

Introduction

Cirrhosis is a disease of the liver and characterized by loss of normal liver function and structure (Rudolph & Kowdley, 1997). Doppler ultrasonography is an important noninvasive diagnostic method in patients with liver disease since cirrhosis and portal hypertension affect the flow profile of the liver vasculature; large portal vein through which oxygen-depleted blood from the stomach, the intestines, the spleen, the gallbladder, and the pancreas flows to the liver. Portal venous blood flow becomes reversed with advanced portal hypertension. Reversed flow is also demonstrated in patients with veno-occlusive disease and portosystemic shunts. Despite general agreement that portal flow velocity is decreased in cirrhotic patients, the absolute values of portal flow velocity in both healthy subjects and cirrhotic patients vary considerably (Kok et al., 1999). Doppler sonography adds information for the evaluation of patients with liver cirrhosis (Bolondi, Gaiani, & Barbara, 1991). Doppler technique is also used to evaluate suspected thrombosis in portal–hepatic vessels and for assessment of transjugular intrahepatic portosystemic stent shunt function (Foshager, Ferral, Nazarian, Castaneda-Zuniga, & Letourneau, 1995). The diagnosis of liver lesions may be improved when Doppler sonography is applied (Forsberg et al., 1995, Leen et al., 1994, Maresca et al., 1994, Nomura et al., 1993, Tanaka et al., 1992). Doppler waveform analysis is often used as a diagnostic tool in the clinical assessment of disease. Spectral analysis is still a preferred method of displaying quantitative Doppler information (Zwiebel, 1987). The use of spectrum analysis to display Doppler frequency shift signals not only provided the best means of measuring blood-flow velocity and also information about the presence of disturbed flow (Green, 1964, Sigel, 1998).

Many spectral estimation methods have recently been developed for Doppler ultrasonic signal processing to diagnostic of valvular diseases (Guo et al., 1994, Matani et al., 1996, Akay and Welkowitz, 1990, Fischer et al., 1998, Barakat, 2002, Hoeks et al., 1993, Hoeks et al., 1991). The results of the studies in the literature have shown that Doppler ultrasound evaluation can give reliable information on both systolic and diastolic blood velocities and have supported that Doppler ultrasound is useful in screening certain hemodynamic alterations in arteries and veins (Sigel, 1998, Evans et al., 1989, Übeyli and Güler, 2004, Müller et al., 2001). Spectral analysis and signal decomposition continue to find wide use in a multitude of biomedical signal processing. Up to now, there is no work relating to the power spectral distribution and a detailed examine documentation about portal vein Doppler spectral waveform changes in the cirrhosis disease in the literature.

In this study Doppler Power spectral density (PSD) estimates of portal venous Doppler signals were obtained by using spectrum analysis techniques. By using spectrum analysis techniques, the variations in the shape of the Doppler PSDs were presented in order to obtain medical information. The aim of this study is evaluate and apply of artificial neural network to power spectral density acquired with Short Time Fourier Transform (STFT) method of portal venous Doppler signals because more confidential method was revealed to diagnose cirrhosis disease. We have implemented an Artificial Neural Network (ANN) that will not only simplify the diagnosis but also enable the physician to make a quicker judgment about the existence of cirrhosis disease more confidencely (Wright, Gough, Rakebrandt, Wahab, & Woodcock, 1997).

Applications of ANNs in the medical field are numerous and include diagnosis of several diseases (Edenbrandt et al., 1993, Chen et al., 1995, Prahadan et al., 1996, Mobley et al., 2000, Abel et al., 1996, Siebler et al., 1994, Smith et al., 1996, Wright and Gough, 1999). These numerous applications exhibit the suitability of ANNs in pattern classification including diagnosis of diseases. A potential application of neural networks is predicting medical outcomes such as diagnostic cirrhosis using portal vein signals.