科技创新与应用2025，Vol.15Issue(22)：27-30,4.DOI:10.19981/j.CN23-1581/G3.2025.22.007
基于YOLOv8算法优化的肺结节检测
摘要
Abstract
At this stage,lung nodule detection has problems such as difficulty in extracting micro-nodule features,poor detection results,complex algorithm models,and large calculation workload.This paper proposes a HS-FPN fusion mechanism based on YOLOv8 to optimize the overall network structure.Add a CA module to the neck of the backbone network for feature extraction to screen lung nodules of different size.After weight calculation,the feature information in each channel can be effectively extracted.ConvTranspose2d convolution is used to scale through bilinear interpolation and transpose convolution to fuse high-level and low-level features of lung nodules,and the CIoU loss function is used to comprehensively evaluate target matching.In the detection of solid lung nodules,the average accuracy reached 86.5%,which was 2.8%higher than the YOLOv8 algorithm.关键词
实性肺结节检测/YOLOv8/HS-FPN/CIoU/微小结节Key words
solid pulmonary nodule detection/YOLOv8/HS-FPN/CIoU/micro nodules分类
信息技术与安全科学引用本文复制引用
杨金全,孙鹏,刘桐泽,牛志恒,陈永君..基于YOLOv8算法优化的肺结节检测[J].科技创新与应用,2025,15(22):27-30,4.基金项目
辽宁科技大学大学生创新创业训练计划项目(202510146011) （202510146011）
