Gastric　cancer　is　one　of　the　most　common　cancers　worldwide.　The　N　staging　of　gastric　cancer　is　a　decisive　factor　in　prognostic　evaluation　and　decision-making　of　staged　cancer　treatment　strategies.　Pathologists　generally　judge　the　N　staging　of　gastric　cancer　based　on　the　number　of　lymph　node　metastasis　on　the　histopathological　WSIs.　Tumor　cells　invading　the　lymph　nodes　are　called　lymph　node　metastasis,　and　we　call　the　invaded　lymph　nodes　positive　lymph　nodes.　The　mainstream　method　for　judging　positive　lymph　nodes　is　still　by　observing　with　the　naked　eye,　which　has　problems　such　as　heavy　workload,　time-consuming,　and　fatigue　easily.　Here,　we　propose　a　deep　learning　framework　for　identifying　the　number　of　lymph　nodes　and　then　determine　whether　each　lymph　node　is　benign　or　malignant.　The　framework　consists　of　a　lymph　node　detection　network　and　a　lymph　node　benign　and　malignant　classification　network.　After　training,　the　detection　network　and　classification　network　achieved　a　recall　rate　of　94%　and　an　accuracy　rate　of　95.7%,　respectively.　Furthermore,　the　framework　maintains　a　0.936　accuracy　rate　in　the　independent　validation　patient　cases.　The　results　demonstrated　that　our　proposed　framework　can　not　only　reduce　pathological　workload　to　a　certain　extent,　but　can　also　assist　in　identifying　positive　lymph　nodes　and　determining　the　N　staging　of　gastric　cancer　in　patients.　©　2021　ACM.