Automatic　segmentation　and　early　diagnosis　of　brain　tumor　is　a　challenging　problem　in　computer　vision　and　it　can　provide　possibility　for　pre-operative　planning,　and　solve　the　problem　such　as　low　accurateness　and　time-consuming　in　traditional　manual　segmentation.　Under　the　mentioned　problems　above,　this　paper　put　forward　a　new　method:　Based　on　traditional　convolutional　neural　networks　(CNNs),　a　new　architecture　model　is　proposed　for　automatic　brain　tumor　segmentation,　which　combines　multi-modality　images.　The　newly　designed　CNNs　model　automatically　learns　useful　features　from　multi-modality　images　to　combine　multi-modality　information.　Experiment　results　show　that　the　proposed　model　is　more　accurate　than　traditional　methods　and　can　provide　reliable　information　for　clinic　treatments.　©　2016　IEEE.