Abstract: Objective: To find the best combination of serum tumor markers to establish a pattern for the diagnosis of lung carcinoma. Methods:The CEA, AFP, CA19-9, CA 72-4, CA 242 , CYFRA 21-1, NSE and TPA levels were detected in 100 lung carcinoma serum samples and 113 healthy serum samples. The samples were divided into two groups. The training group contained 150 samples (70patients and 80healthy people) and the test group contained 63samples (30patients and 33healthy people). We evaluated the serum tumor markers with the area under curves, selected the optimum serum tumor marker combination and built the diagnostic pattern with an artificial neural network. Results: CA211 and CEA were selected to be the optimum serum tumor marker combination and the artificial neural network was built based on these markers. This diagnostic lung carcinoma pattern has a specificity of 92.9% , sensitivity of 86.0% and positive value of 85.5%.Conclusion:This combination of optimum serum tumor markers has established a pattern with high sensitivity and specificity for the detection of lung carcinoma. It has the potential to become a valuable clinical tool for early diagnosis.