This paper describes the medical information retrieval systems designed by the Poznan University of Technology of for clinical decision support CDS) which were submitted to the TREC 2016. The baseline is the Terrier DPH Bo1 which recently has been shown to be the most effective Terrier option. We also used Mesh query expansion, word2vec query expansion, and the combination of these two options. In all measures our results are approximately 0,02 above the median.