| 76 | 0 | 10 |
| 下载次数 | 被引频次 | 阅读次数 |
针对儿童腕部X光图像中细微骨折检测精度较低的问题，提出一种基于YOLOv11n改进的MSEF-YOLO(multi-scale efficient fusion network-YOLO)目标检测算法。首先，将空间和通道重构卷积(SCConv)模块与C3k2模块融合，通过空间重构单元(SRU)和通道重构单元(CRU)并行处理空间与通道的冗余，增强对小目标的感知能力；其次，引入多尺度扩张注意力(MSDA)机制提高特征提取能力，进而提高模型检测精度与泛化性，有效减少漏检和误检；最后，优化尺度序列特征融合(SSFF)模块并设计SSFF-X模块，通过3D卷积增强多尺度特征融合能力，进一步提升对细微骨折的检测效果。实验结果表明，相较于原YOLOv11n算法，MSEF-YOLO算法的精确率、召回率、mAP@0.5和mAP@0.5～0.95分别提高了3.1%、3.8%、3.0%和3.5%。MSEF-YOLO算法能够有效协助放射科医生检测儿童腕部骨折，为医学图像的诊断提供技术支持。
Abstract：To address the issue of low detection accuracy for subtle fractures in pediatric wrist X-ray images, an improved YOLOv11n-based object detection algorithm called MSEF-YOLO(multi-scale efficient fusion network-YOLO)is proposed.First, the spatial and channel reconstruction convolution(SCConv)module is integrated with the C3k2 module, utilizing the spatial reconstruction unit(SRU)and channel reconstruction unit(CRU)to process spatial and channel redundancy in parallel, thereby enhancing the perception of small objects.Second, the multi-scale dilated attention(MSDA)mechanism is introduced to improve feature extraction capability, thereby enhancing detection accuracy and generalization ability while effectively reducing missed and false detections.Finally, the scale sequence feature fusion(SSFF)module is optimized and the SSFF-X module is designed, leveraging 3D convolution to enhance multi-scale feature fusion, further improving the detection performance for subtle fractures.Experimental results demonstrate that, compared to the original YOLOv11n algorithm, the MSEF-YOLO algorithm improves precision, recall, mAP@0.5,and mAP@0.5～0.95 by 3.1%,3.8%,3.0%,and 3.5%,respectively.The MSEF-YOLO algorithm effectively assists radiologists in detecting pediatric wrist fractures, providing technical support for medical image diagnosis.
[1] THAIYALNAYAKI K,KAVYAA L,SUGUMAR J.Automated bone fracture detection using convolutional neural network[J].Journal of Physics:Conference Series,2023,2471(1):012003.
[2] THIAN Y L,LI Y T,JAGMOHAN P,et al.Convolutional neural networks for automated fracture detection and localization on wrist radiographs[J].Radiology:Artificial Intelligence,2019,1(1):e180001.
[3] ZOU J T,ARSHAD M R.Detection of whole body bone fractures based on improved YOLOv7[J].Biomedical Signal Processing and Control,2024,91:105995.
[4] WU H T,FU Q S.An innovative deep learning approach to spinal fracture detection in CT images[J].Annali Italiani Di Chirurgia,2024,95(4):657-668.
[5] 周涛，杜玉虎，茆晶晶，等.面向下颌骨骨折检测的主辅网络YOLOv5模型[J].中国科技论文，2023,18(11):1257-1266.ZHOU T,DU Y H,MAO J J,et al.YOLOv5 model of primary and auxiliary networks for mandibular fracture detection[J].China Sciencepaper,2023,18(11):1257-1266.(in Chinese)
[6] ZHOU T,WANG H W,DU Y H,et al.M3YOLOv5:feature enhanced YOLOv5 model for mandibular fracture detection[J].Computers in Biology and Medicine,2024,173:108291.
[7] LIU X H,ZHOU Y H,DONG Z H,et al.LSDASCFormer:a transformer-like network with long-short-distance attention and SCConv for heart rate measurement[J].IEEE Sensors Journal,2024,24(17):28007-28017.
[8] JIAO J Y,TANG Y M,LIN K Y,et al.DilateFormer:multi-scale dilated transformer for visual recognition[J].IEEE Transactions on Multimedia,2023,25:8906-8919.
[9] 王震洲，杨榕，宿景芳.改进YOLOv8的烟盒缺陷检测算法[J].电子测量技术，2024,47(13):110-119.WANG Z Z,YANG R,SU J F.Improved YOLOv8 cigarette box defect detection algorithm[J].Electronic Measurement Technology,2024,47(13):110-119.(in Chinese)
[10] 蒋仁杰，袁贞明，吴英飞.HR-SCNet在儿童发育性髋关节发育不良诊断中的应用研究[J].中国数字医学，2024,19(10):1-7.JIANG R J,YUAN Z M,WU Y F.Application of HR-SCNet in the diagnosis of developmental dysplasia of the hip in children[J].China Digital Medicine,2024,19(10):1-7.(in Chinese)
[11] 冯迎宾，刘艾妮.基于YOLOv8n的夜间车辆检测[J].沈阳理工大学学报，2025,44(2):1-6,12.FENG Y B,LIU A N.Night-time vehicle detection based on YOLOv8n[J].Journal of Shenyang Ligong University,2025,44(2):1-6,12.(in Chinese)
[12] 周秦汉，贾杰，陈昊，等.基于SKNet改进YOLOv5s的无人机对道路小目标的检测[J].南昌航空大学学报(自然科学版),2023,37(4):39-45.ZHOU Q H,JIA J,CHEN H,et al.Improving YOLOv5s based on SKNet for unmanned aerial vehicle detection of small road targets[J].Journal of Nanchang Hangkong University(Natural Sciences),2023,37(4):39-45.(in Chinese)
[13] 厍向阳，景任杰，董立红.3D密集全卷积的高光谱地物分类算法研究[J].计算机工程与应用，2023,59(3):112-117.SHE X Y,JING R J,DONG L H.Research on hyperspectral ground object classification algorithm based on 3D dense full convolution[J].Computer Engineering and Applications,2023,59(3):112-117.(in Chinese)
[14] NAGY E,JANISCH M,HR,et al.A pediatric wrist trauma X-ray dataset(GRAZPEDWRI-DX)for machine learning[J].Scientific Data,2022,9(1):222.
[15] WAN D H,LU R S,SHEN S Y,et al.Mixed local channel attention for object detection[J].Engineering Applications of Artificial Intelligence,2023,123:106442.
[16] WANG Y Q,ZHANG L Y,XIE J J,et al.DFEFM:fusing frequency correlation and mel features for robust edge bird audio detection[J].Avian Research,2025,16(2):100232.
基本信息:
DOI：
中图分类号:R726.8;TP391.41
引用信息:
[1]宫硕,蒋强,李婷雪.基于MSEF-YOLO的儿童腕部骨折检测算法[J].沈阳理工大学学报,2025,44(05):29-36.
基金信息:
辽宁省教育厅高等学校基本科研项目(LJ21241014452)