The　examination　of　the　lungs　is　an　important　part　of　the　annual　physical　examination.　There　are　hundreds　or　thousands　of　cases　in　the　physical　examination,　and　each　case　contains　many　lung　cross-sectional　CT　images.　These　all　require　professional　doctors　to　screen　for　cases　with　pulmonary　nodules　one　by　one,　which　is　not　only　a　heavy　workload　but　also　a　possibility　of　incorrect　screening.　Aiming　at　the　above　problems,　a　Convolutional　Neural　Network　(CNN)　is　introduced　to　screen　out　the　CT　images　for　pulmonary　nodules,　and　a　classification　algorithm　based　on　CNN　is　proposed.　The　experimental　results　in　the　LIDC　database　show　that　compared　with　the　widely　used　lenet-5　network,　traditional　methods,　and　other　deep　learning　models,　the　use　of　customized　convolutional　neural　networks　improves　the　classification　accuracy.　The　AUC　value　is　0.821　6,　which　is　also　the　highest　among　several　classifiers.　Compared　with　other　methods,　this　method　can　more　accurately　identify　CT　images　of　the　lungs　and　can　provide　a　more　objective　reference　for　clinical　diagnosis　as　it　can　be　used　for　CAD　systems.　©　2020　ACM.