In　this　paper,　we　propose　a　diagnosis　and　classification　method　of　colonography　CT　Images　based　on　deep　learning　and　image　reconstruction,　which　is　processed　to　increase　the　difference　between　pathological　features　and　other　healthy　tissues　and　improve　the　accuracy　of　pathological　images　classification.　It　is　difficult　to　accurately　distinguish　polyps　features　owing　to　the　variability　of　appearances,　fuzzy　boundaries,　heterogeneous　densities,　shapes　and　sizes　of　lesions.　Some　study　results　show　that　the　ResNet　network　has　a　better　performance　than　other　methods　and　has　broad　application　prospects　in　the　identification　of　CT　images.　We　used　the　improved　ResNet　network　as　the　classification　model.　We　completed　the　classification　experiment　on　the　colonography　datasets.　Through　comparative　experiments,　we　found　gradient　features　play　an　important　role　in　the　classification　of　colonography　CT　images.　More　unexpectedly,　for　CT　images　of　colonography,　the　classification　effect　of　CT　images　with　small　information　entropy　is　excellent.　A　new　＇RGB＇　image　containing　two　channels　of　gradient　features　and　one　channel　of　LBP　features　is　very　effective　for　classification,　which　may　be　a　new　attempt.　©　2021,　The　Author(s),　under　exclusive　license　to　Springer　Nature　Singapore　Pte　Ltd.