融合注意力机制的3DV-Net肺结节检测

打开文本图片集

中图分类号：TP391 文献标志码：A 文章编号：1007-2683（2025）04-0123-11

3D V-Net Detection of Pulmonary Nodules with Integrated Attention Mechanisms

WEI Haiyue， CHEN Hailong， XU Xinyao， ZHANG Xiuxia， ZHOU Xinpeng （Schoolof ComputerScienceand Technology，Harbin Universityof Science and Technology，Harbin15Oo80，China）

Abstract：Pulmonary nodulesplayacrucialroleintheearlydiagnosisandtreatment of lungcancer.Toadresstheexistingissues of misedandfalsedetectionsincurent pulmonary noduledetectionalgorithms，weproposea3DV-Netmodelthatincorporatesan atentionmechanismforimprovedpulmonarynoduledetection.Initiall，wedevelopedthe3DV-Netmodelforbasicpulmonarynodule detection，ensuring highaccuracyinidentifyingnodules.Subsequently，weemployedtheConvolutionalBlock AttentionModule （CBAM）toehancethefeaturesofthegeneratedimagesbyconsideringbothchanelandspatialinformation，therebyimprovingmodel performance.Finally，wetrainedandevaluatedthemodelusing thepubliclyavailableLUNA16dataset.Theexperimentalresultsshow thatthe proposed3DV-Netmodel，whichintegratesteaentionmechanism，improvesthedetectionperformanceinpulmonaryodule detection.The intersection over union （IoU）between the predicted results and the true labels reached 84.88% ，and the Dice loss achieved -90.43% ，demonstrating the feasibility of the model and the accuracy of the detection.

Keywords：pulmonary nodule detection;deep learning;medical image processing；3DV-Net network；atention mechanism

0 引言

肺癌是全球范围内发病率和致死率最高的癌症之一[1]，肺结节作为肺癌早期诊断和治疗的重要标志物，对于提供准确的诊断和个体化的治疗方案具有关键性作用[2]。（剩余13882字）