基于改進(jìn)YOL0v8s的肺結(jié)節(jié)檢測(cè)算法
打開(kāi)文本圖片集
中圖分類號(hào)：TP391.4；TP183 文獻(xiàn)標(biāo)識(shí)碼：A 文章編號(hào)：2096-4706（2025）07-0087-06
Abstract：Lung cancer has become one of the common cancers with high mortality worldwide.Pulmonary nodules are the early manifestationsoflungcancer.Aimingatthediffcultyof pulmonarynoduledetectioninCTimages，apulmonary nodule detection algorithm based on improved YOLOv8s is proposed.The backbone network uses improved YOLOv8s，and uses Space-to-Depth Convolution toreplacethe traditional step-sizeconvolutionand poling layertoavoid thelossoffine-grained information caused by step-size convolution and poling layer when processing low-resolution images or small objects.The coordinateattentionmoduleisaddedtoconsiderteinter-channelrelationshipandpositioninformationofthepulmonarynodule image，othat themodelcanlocateandidentifythetargetareamoreaccurately.FocalLossisusedtoreplacethecros-entropy loss function to solvethe problemofsmallnumberofsample labels inthe dataset.Theadaptive activationfunctioncan not only improve the stabilityofthe network，butalso improve theaccuracyofthe network.The LUNA16datasetisusedtoverifythe performance of the algorithm.The detection accuracy of the improved pulmonary nodule detection algorithm reaches 7 7 . 8 % ， which is 3 . 7 % and 8 . 6 % higher than that ofFasterR-CNNand YOLOv8sdetectionalgorithms，respectively.
Keywords： pulmonary nodule detection;YOLOv8s;no more strided convolution;adaptiveactivation function; spatial attentionmechanism
0 引言
肺癌已成為全球常見(jiàn)癌癥中死亡率最高的疾病[]。（剩余6753字）