Objective: To evaluate the feasibility and the efficacy of medical diagnosis expert system, combined with preoperative magnetic resonance cholangiopancrcatography(MRCP), in predicting common bile duct(CBD) stones. 
Methods: The clinical data of 286 cases of cholecystolithiasis and cholecystitis admitted to our hospital during Jan 2009 to May 2009 were analysed prospectively. These cases were divided randomly into the control group (n=143) and the research group (n=143). The control group underwent routine MRCP before cholecystectomy, while the research group was screened by medical diagnosis expert system which had been established based on artificial neural network technique and proved to be stable. MRCP was performed in every patient of the research group when the screening result was positive. Then, the positive rate of MRCP, the mean hospitalization time and the average expenditure between this two groups were compared.
Results: CBD positive rate was 16.78%(24/143) of the control group, while the research group was 16.08% (23/143). The positive rate of MRCP of the research group was 95.24%（20/21）,significantly higher than 14.69%（21/143）of the control group（P＜0.01）. The mean hospitalization time of research group (8.83±3.59 d) was significantly shorter than that of the control group (10.10±3.66 d)  (P＜0.01).The average expenditure of the research group was (11 175.80±7 972.33) Yuan, evidently lower than (13 332.08±9 884.04) Yuan of the control group (P＜0.05). 
Conclusions: The medical diagnosis expert system has a remarkable performance in predicting CBD stones. It can be used routinely to select the appropriate cases to undergo MRCP, which could obviously reduce the hospitalization costs and shorten the hospitalization time.