MEDICAL IMAGE RECOGNITION OF ABDOMINAL MULTI-ORGANS BY RBF GMDH-TYPE NEURAL NETWORK

作者：Kondo Tadashi^*; Ueno Junji

来源：International Journal of Innovative Computing Information and Control, 2009, 5(1): 225-240.

• Neural network
• GMDH
• Medical image recognition

In this paper, a Radial Basis RBF) Group Method of Data Handling (GMDH)-type neural network algorithm is proposed and applied to the medical image recognition of abdominal X-ray CT images. In the RBF GMDH-type neural network, two kinds of neuron architectures such as RBF type neuron and polynomial type neuron, are used for self-organizing the optimum neural network architecture and the structural parameters such as the number of layers, the number of neurons in the hidden layers, the optimum neuron architecture and useful input variables are automatically determined so as to minimize the prediction error criterion defined as Prediction Sum of Squares (PSS) by the heuristic self-organization method. In the medical image recognition of the abdominal X-ray CT images, the image densities of the abdominal organs such as the liver, stomach and spleen are very similar each other, and so it is very difficult to recognize the regions of these organs accurately. In this paper, the RBF GMDH-type neural network is applied to the medical image recognition of abdominal organs. The optimum neural networks fitting the image characteristics of these organs art! automatically organized using RBF GMDH-type neural network algorithm, and the regions of the abdominal organs are recognized and extracted accurately. The recognition results are compared with those obtained using the sigmoid function type neural network network using the back propagation algorithm.

• 出版日期2009-1