简介:Pulmonaryvesselsextractionisachallengingtaskinclinicalmedicine.Manypulmonarydiseasesareaccompaniedbythechangesofvesseldiameters.Thevesselsandtheirbranches,whichexhibitmuchvariability,aremostimportantinperformingdiagnosisandplanningthefollow-uptherapies.Inthispaper,weproposeanefficientapproachtopulmonaryvesselsextractionbasedonthecurveevolution.Thisapproachmodelsthevesselsasmonotonicallymarchingfrontunderthespeedfieldintegratingboththeregionandtheedgeinformationwhereanewregionspeedfunctionisdesignedandintegratedwiththeedgebasedspeedfunction.Duetotheregionbasedspeedterm,thefrontcouldevenpropagateinsmallnarrowvesselbranches.Tofurtherimprovethesegmentationresults,amulti-initialfastmarchingalgorithmisdevelopedtofastimplementthenumericalsolution,whichmayavoidthemonotonicallymarchingfrontleakingoutoftheweakboundarytooearlierandalsoreducethecomputationalcost.ThevalidityofourapproachisdemonstratedbyCTpulmonaryvesselsextraction.Experimentsshowthatthesegmentationresultsbyourapproach,especiallyonthenarrowthinvesselbranchesextraction,aremoreprecisethanthatoftheexistingmethod.
简介:Inthispaper,weproposeanovelautomaticobjectextractionalgorithm,namedtheTemplateGuidedLiveWire,basedonthepopularlyusedlivewiretechniques.Wediscussindetailsthenovelmethod'sapplicationsontongueextractionindigitalimages.Withtheguidesofagiventemplatecurvewhichapproximatesthetongue'sshape,ourmethodcanfinishtheextractionoftonguewithoutanyhumanintervention.Inthepaper,wealsodiscussedindetailshowthetemplateguidesthelivewire,andwhyourmethodfunctionsmoreeffectivelythanotherboundarybasedsegmentationmethodsespeciallythesnakealgorithm.Experimentalresultsonsometongueimagesareaswellprovidedtoshowourmethod'sbetteraccuracyandrobustnessthanthesnakealgorithm.