Medical Image Diagnosis of Liver Cancer by Multi-layered GMDH-type Neural Network Using Principal Component-Regression Analysis and PSS Criterion
Tadashi Kondo, Junji Ueno, Shoichiro Takao · Proceedings of the ISCIE International Symposium on Stochastic Systems Theory and its Applications · 2013
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, heuristic self-organization method, which is a kind of evolutionary computation, is used to organize the neural network architecture. But, multi-colinearity occurs and prediction values become unstable. In this study, hybrid multi-layered GMDH-type neural network using principal component-regression analysis is proposed. In this algorithm, multi-colinearity does not occur and accurate prediction values are obtained. This new algorithm is applied to the medical image diagnosis of liver cancer. First, the GMDH-type neural network which recognizes the liver regions, is automatically organized using multi-detector row CT (MDCT) images of the liver, and the liver regions are recognized and extracted. Then, new another GMDH-type neural network is automatically organized using the extracted image of liver, and the candidate regions of the liver cancer is recognized and extracted. The recognition results are compared with the conventional sigmoid function neural network trained using back propagation method and it is shown that this algorithm is useful for CAD of liver cancer.