To　overcome　the　difficulty　caused　by　the　overlapping　of　cells　in　the　image　analysis　of　tumor　FISH　cells　for　cancer　diagnosis,　a　new　edge　detection　of　tumor　FISH　cells　is　proposed　based　on　automatic　random　walk　algorithm　to　separate　the　clustering　cells.　Firstly,　the　regions　of　the　cell＇s　seeds　are　detected　by　the　ultra-erosion　based　on　priori　knowledge.　Then,　the　effective　pixel　markers　are　extracted　from　the　detected　regions　and　treated　as　the　object　seeds　in　the　random　walk　algorithm.　Finally,　the　automatic　edge　detection　of　the　tumor　FISH　cells　is　realized　by　using　seeds　to　segment　multiple　cells　in　a　random　walk　way.　©　2012　IEEE.