抄録
A new diagnosis support system of chronic liver disease had been proposed. This system is composed of two steps. One is an artificial neural network step. A back propagation neural network was designed to diagnose seven classifications of chronic liver diseases: CIH, CAH, LC, HCC, FL, ALD, and Normal. The other is a medical rule base step. The output of a neural network are checked by this rule base. The network architecture had constructed each disease. Network inputs are consisted within 26 laboratory data per 187 patient cases. After being trained the system classified chronic liver diseases correctly in 141 of 187 cases. An accuracy of 75.4% is higher than the average score (63.0%) by 5 hepatologists. When sufficiently sophisticated, this system may significantly improve the analysis of chronic liver disease.
