华侨大学学报（自然科学版）2025，Vol.46Issue(5)：493-504,12.DOI:10.11830/ISSN.1000-5013.202503008
边缘细节增强的肺炎胸部X射线病灶定位
Edge-Enhanced Lesion Localization in Chest X-Rays for Pneumonia Detection
摘要
Abstract
An improved YOLO11s-SAD algorithm based on YOLO11 is designed to relieve the situation of difficult detection of tiny lesions,poor lesion localization performance in complex backgrounds,and issues of missed and false detections.First,a spatial edge information fusion(SEIF)module is designed,which en-hances the feature extraction capability of backbone for lesion edges by procesings input images in parallel use-ing Sobel operator-based edge detection and max pooling.Then,ASF-Neck is employed as the new neck net-work to better capture relationships between multi-scale features by optimizing the feature fusion mechanism.Finally,dynamic upsampling(DySample)replaces bilinear interpolation in the scale sequence feature fusion(SSFF)module of ASF-Neck to reduce the loss of pneumonia detail features during upsampling.The model parameters are optimized using the Adam optimizer.Experimental results show that the proposed algorithm a-chieves a mean average precision of 57.9％,improving by 3.4％compared to the baseline,while introducing no significant increase in parameters or floating-point operations.The lesion localization performance also out-performs other mainstream detection algorithms.关键词
肺炎检测/病灶定位/辅助诊断/SEIF模块/YOLO11算法Key words
pneumonia detection/lesion localization/aided diagnosis/SEIF module/YOLO11 algorithm分类
计算机与自动化引用本文复制引用
臧佳明,郑力新,何建海,潘书万..边缘细节增强的肺炎胸部X射线病灶定位[J].华侨大学学报（自然科学版）,2025,46(5):493-504,12.基金项目
福建省科技计划项目(2020Y0039) （2020Y0039）
