Hybrid Multi-layered GMDH-type Neural Network Using Principal Component Regression Analysis and its Application to Medical Image Diagnosis of Liver Cancer

Abstract

In this study, a hybrid multi-layered Group Method of Data Handling (GMDH)-type neural network algorithm using principal component-regression analysis is proposed and applied to the computer aided image diagnosis (CAD) of liver cancer. In the GMDH-type neural network, a heuristic self-organization method that is a type of evolutionary computation, is used to organize the neural network architecture. In this revised GMDH-type neural network, the optimum neural network architecture is automatically organized from three types of neural network architectures, such as the sigmoid function neural network, the radial basis function (RBF) network and the polynomial neural network architecture, by the heuristic self-organization method. Furthermore, the structural parameters such as the number of layers, the number of neurons in hidden layers and useful input variables, are automatically determined using the heuristic self-organization method. In the revised GMDH-type neural network proposed in this paper, the principal component-regression analysis is used to protect multi-colinearity which has occurred in the learning calculations of neurons, and accurate and stable prediction values are obtained. This new algorithm is applied to the medical image diagnosis of liver cancer. In this application, two types of neural network architectures fitting the complexity of the multi-detector row CT (MDCT) medical images, are automatically organized using the revised GMDH-type neural network algorithm The first neural network recognizes and extracts the liver regions from the MDCT images of the liver, and the second neural network recognizes and extracts the liver cancer regions. These results are compared with the conventional sigmoid function neural network trained using the back propagation method, and this GMDH-type neural network algorithm is shown to be useful for CAD of liver cancer.