PubMed, freely available on the internet, is the best known database for medical information. We propose a method of optimization of the PubMed Automatic Term Mapping (ATM) that includes MeSH terms. This method is evaluated using two queries constructed to emphasize the differences between the PubMed queries as they are at present and also between these queries and the optimized one. The proposed query is significantly more precise than the current PubMed query (54.5% vs. 27%). The optimized query proposed would be easy to implement into PubMed.