Radioactive　seed　implantation　has　emerged　as　an　effective　treatment　modality　where　small　radioactive　seeds　are　implanted　into　the　target　organ　to　eradicate　the　cancer　by　emitting　radiation.　Precise　seeds　localization　can　indicate　whether　those　seeds　deliver　sufficient　doses　of　radiation.　However,　it　is　challenging　and　laborious　to　identify　all　seeds　manually　in　a　short　time.　Therefore,　our　purpose　in　this　study　was　to　develop　an　automatic　technique　for　identifying　implanted　seeds　on　any　parts　of　body.　The　algorithm　relies　on　a　3D　adaptive　median　filter　to　remove　bone　structure;　white　top-hat　transform　to　extract　seeds-like　objects　and　further　seed　classification　analysis　based　on　size,　shape　and　their　connection　etc.　Preliminary　results　on　ten　patients　and　seven　simulated　data　show　that　this　approach　to　be　effective　and　accurate.　It　resulted　in　a　96.9　%　detection　rate　with　a　corresponding　4.7　%　false-positive　rate　for　clinical　data;　a　98.5　%　detection　rate　with　a　corresponding　4.1　%　false-positive　rate　for　the　simulated　data;　and　sub-millimeter　accuracy　for　both　data　sets.　This　method　can　achieve　robust　and　accurate　seed　segmentation　through　the　proposed　workflow.