Statistical Neural Networks in the Classification of Alcoholic Liver Disease and Nonalcoholic Fatty Liver Disease

Pwasong A Davou,Saratha Sathasivam

doi:10.26706/ijceae.1.1.20150102

Abstract

This paper deals with the performance of statistical neural network in the classification of alcoholic liver disease (ALD) data and nonalcoholic fatty liver disease data (NAFLD). The study involved 73 individuals that were clinically diagnosed of alcoholic liver disease (ALD) and 80 individuals who were clinically diagnosed of nonalcoholic fatty liver disease (NAFLD). Four different neural network structure, multi-layer perceptron, radial basis function, probabilistic neural network and generalized regression neural network were applied to the data to determine the performance of statistical neural networks in the classification of liver disease data. The overall result indicates that the most suitable statistical neural network model for classifying ALD and NAFLD data is the probabilistic neural network (PNN) with a 95.7% classification performance and 67 correct classifications. Radial basis function network (RBF) and multilayer perceptron network (MLP) has the lowest classification accuracy with 55 classified samples each. The generalized regression neural network (GRNN) was the second-best network with 62 correct classifications. The computer simulation was carried out by using MATLAB 6.0 Neural Network Toolbox.

Highlights

IntroductionCirrhosis is the final phase of alcoholic liver disease
This study illustrated the manner statistical neural networks are used in actual clinical diagnosis of alcoholic liver disease and nonalcoholic fatty liver disease
By applying statistical neural networks, a diagnostic classification system that performs at an accuracy level is constructed here

Summary

Introduction

Cirrhosis is the final phase of alcoholic liver disease. It is an information handling pattern that is inspired by the way the brain processes information at the low biological level It look like the brain in terms of knowledge being developed by the network through a learning process and inter-neuron linking strengths recognized as synaptic weights which are used to store the knowledge. Pye and Bangham [5] assert that when it is determined that an object from a population p belongs to a known subpopulation s, it is said that pattern recognition is done He further stated that the recognition of an individual object as a unique singleton class is called identification. Compared with other approaches; their algorithm is a precise artificial intelligent approach which is fast in performance and easy in enactment

Methods

Findings

Discussion

Conclusion