In our previous study, we attempted to develop a tool for the early diagnosis of non-small cell lung cancer (NSCLC) using DNA methylation biomarkers. With the aim of improving the diagnostic potential by optimizing the composition of the target set, in this study, 13 candidate genes (ACTA1, AIDH1A2, CBX8, CDH8, EVX1, MGC16275, NEUROG1, NEUROG2, NID2, OTX2OS1, PGAM2, PHOX2B and TOX) were analyzed by methylation-specific PCR to determine the methylation status of each gene in 5 NSCLC cell lines and in lung tissue samples from 15 healthy volunteers, 103 stage I NSCLC patients and 26 non-cancerous control patients. Results showed that NEUROG2 and NID2 were hypermethylated in stage I NSCLC tissues (31.07 and 46.60%, respectively) and unmethylated in normal lung tissues (0/15) and non-cancerous tissues (0/26). Following recombination, an optimized 5-gene panel (NEUROG2, NID2, RASSF1A, APC and HOXC9) achieved a sensitivity of 91.26% with a specificity of 84.62% in the detection of stage I NSCLC. The optimized 5-gene panel greatly improved the diagnostic power for stage I NSCLC.