Due　to　the　irregular　and　manifold　shapes　of　brain　tumors,　it　is　somewhat　complicate　to　match　the　tumor　images　to　the　right　result　and　provide　the　guidance　to　assess　and　improve　the　quality　as　well　as　the　improvement　of　individual＇s　actions.　Traditional　methods　using　simple　method　to　deal　with　the　problem.　The　result　shows　that　it　is　time-consuming　and　hard　to　put　into　practice　in　clinical　application.　In　this　paper,　this　study　made　the　attempt　to　improve　the　algorithm　and　it　made　sense.　In　this　work,　an　efficient　algorithm　combining　3　methods　was　proposed　and　a　comparison　was　made　to　their　performance.　The　new　algorithm　briefly　makes　a　lower　complexity　of　the　layer　function.　Experimental　results　apparently　shows　that　our　algorithm　is　rather　competitive.　It　can　provide　accurate　feedback　for　learners　in　brain　tumor　recognition.　The　precision　of　proposed　CNN　can　reach　0.69　while　MobileNet　method　and　VGG16　can　reach　0.75　and　0.71.　Moreover,　the　low　rate　of　loss　makes　our　model　much　more　stable.　The　satisfied　result　makes　our　method　a　remarkably　promising　tool　in　medical　treatment.　©　2022　IEEE.