With　the　explosive　growth　of　medical　information,　the　users　not　only　query　information　efficiently　and　accurately,　but　also　pay　attention　to　the　information　of　sensitivity　and　privacy.　In　medical　domains,　the　current　privacy　preserving　methods　either　use　the　technology　of　Access　Control　List,　or　need　to　prepare　training　documents　for　each　privacy　policy.　However,　it　is　a　time-consuming　and　impractical　way　for　data　owners　to　assign　a　privacy　policy　on　each　document.　In　this　paper,　by　exploiting　the　privacy　medical　queries,　we　propose　a　novel　approach　based　on　semantic　and　ontology　to　achieve　the　topic-level　privacy　preserving　search.　With　the　support　of　them,　we　first　mine　all　the　potential　hierarchy　and　semantics　from　a　user　query　and　acquire　sensitive　terms　relative　to　privacy　policies　automatically　without　training　documents.