
浏览全部资源
扫码关注微信
1.陆军军医大学(第三军医大学)第二附属医院超声科,重庆
2.陆军军医大学(第三军医大学)生物医学工程与影像医学系数字医学教研室,重庆
徐亚丽,E-mail:xuyal1976@tmmu.edu.cn
收稿:2025-12-18,
修回:2026-01-13,
录用:2026-01-19,
纸质出版:2026-02-28
移动端阅览
李苇,李华龙,曾小科,等.一种可显著提升初级超声医师诊断准确性与效率的深度学习模型:基于YOLOv13的腹部超声异常实时识别系统开发与验证[J].陆军军医大学学报,2026,48(04):493-503.DOI:10.16016/j.2097-0927.202512098.
LI W,LI H L,ZENG X K,et al.A deep learning-based model for significantly enhancing diagnostic accuracy and efficiency in novice sonographers: development and validation of a YOLOv13-powered real-time abdominal ultrasound anomaly detection system[J].J Army Med Univ,2026,48(04):493-503.DOI:10.16016/j.2097-0927.202512098.
李苇,李华龙,曾小科,等.一种可显著提升初级超声医师诊断准确性与效率的深度学习模型:基于YOLOv13的腹部超声异常实时识别系统开发与验证[J].陆军军医大学学报,2026,48(04):493-503.DOI:10.16016/j.2097-0927.202512098. DOI:
LI W,LI H L,ZENG X K,et al.A deep learning-based model for significantly enhancing diagnostic accuracy and efficiency in novice sonographers: development and validation of a YOLOv13-powered real-time abdominal ultrasound anomaly detection system[J].J Army Med Univ,2026,48(04):493-503.DOI:10.16016/j.2097-0927.202512098. DOI:
目的
2
开发并验证一种基于腹部超声影像的深度学习模型,用于17类常见腹部异常的实时识别与定位,为基层超声诊断提供可靠技术支撑。
方法
2
回顾性收集陆军军医大学第二附属医院超声科2017年1月至2024年9月腹部异常患者18 947例共39 530幅B型超声图像。按异常类别分层随机抽取1 891幅图像构建独立测试集;剩余的37 639幅图像按7∶1∶2比例随机划分为训练集(26 258幅)、验证集(3 840幅)及内部测试集(7 541幅)。经标准化标注后,在内部测试集上对比RetinaNet、Faster R-CNN、RT-DETR及YOLOv13这4种主流目标检测网络的性能,筛选最优模型;评估指标包括交并比(intersection over union, IoU)、精确率(precision)、召回率(recall)、F
1
值(F
1
)及平均精度均值(mAP@50);在独立测试集上,进一步比较4种诊断模式的诊断效能(最优模型、3名无标注经验的初级超声医师、 3名中级超声医师及模型辅助下初级医师)。
结果
2
内部测试集上,YOLOv13模型表现最优,mAP@50为0.653(95%
CI
:0.608~0.669),高于RetinaNet(0.546,95%
CI
:0.511~0.563)、Faster R-CNN(0.578,95%
CI
:0.531~0.591)及RT-DETR(0.627,95%
CI
:0.585~0.643);其精确率为0.633(95%
CI
:0.624~0.643),召回率为0.715(95%
CI
:0.704~0.725),F
1
值为0.671(95%
CI
:0.663~0.681);独立测试集上,IoU为0.5时,6名医师的F
1
值均值为0.57(初级组:0.484,中级组:0.656);YOLOv13模型的F
1
值为0.624,显著高于所有初级医师(
P
<
0.001)。模型辅助后,初级医师的F
1
值均值从0.484提升至0.644(提升33.1%,
P
<
0.001),接近中级医师水平;平均精确率从0.529提升至0.638(提升20.6%),平均召回率从0.472提升至0.674(提升42.8%),平均诊断耗时从213.33
min缩短至161.33 min(缩短24.4%)。
结论
2
YOLOv13深度学习模型在17类腹部异常超声检测中表现最优,能显著提升初级超声医师的诊断效能,可作为基层及社区医院腹部超声筛查的辅助工具。
Objective
2
To develop and validate a deep learning model for real-time identification and localization of 17 common abdominal anomalies on ultrasound images
providing reliable technical support for primary care sonography.
Methods
2
Retrospectively collected 39 530 B-mode ultrasound images from 18 947 patients with abdominal anomalies at the Second Affiliated Hospital of Army Medical University from January 2017 to September 2024). A stratified random sample of 1 891 images formed an independent testing set; the remaining 37 639 images were split 7:1:2 into training (26 258)
validation (3 840)
and internal test (7 541) sets. After standardized annotation
RetinaNet
Faster R-CNN
RT-DETR
and YOLOv13 were compared on the internal test set using IoU
precision
recall
F
1
-score
and mAP@50. The optimal model was further evaluated against 6 sonographers (3 novice
3 intermediate) and novice sonographers assisted by the model on the independent test set.
Results
2
YOLOv13 achieved superior performance on the internal testing set: mAP@50 0.653 (95%
CI
:0.608 to 0.669) versus RetinaNet 0.546 (95%
CI
:0.511 to 0.563)
Faster R-CNN 0.578 (95%
CI
:0.531 to 0.591)
RT-DETR 0.627 (95%
CI
:0.585 to 0.643); precision 0.633 (95%
CI
:0.624 to 0.643)
recall 0.715 (95%
CI
:0.704 to 0.725)
F
1
-score 0.671 (95%
CI
:0.663 to 0.681). On the independent testing set (IoU=0.5)
YOLOv13’s F
1
-score (0.624) exceeded all novice sonographers’ mean F
1
-score (0.484;
P
<
0.001). Model assistance improved novice sonographers’ mean F1-score from 0.484 to 0.644 (33.1% improvement;
P
<
0.001)
approaching intermediate-level performance (0.656); mean precision increased 20.6% (0.529 to
o
0.638)
mean recall increased 42.8% (0.472 to 0.674)
and mean interpretation time decreased 24.4% (213.33 min to 161.33 min).
Conclusion
2
The YOLOv13 model demonstrates optimal performance for detecting 17 abdominal ultrasound anomalies and significantly enhances diagnostic efficiency of novice sonographers
serving as an effective assistive tool for abdominal ultrasound screening in primary care settings.
TOMIZAWA M , SHINOZAKI F , HASEGAWA R , et al . Abdominal ultrasonography for patients with abdominal pain as a first-line diagnostic imaging modality [J]. Exp Ther Med , 2017 , 13 ( 5 ): 1932 - 1936 . DOI: 10.3892/etm.2017.4209 http://dx.doi.org/10.3892/etm.2017.4209 .
孙蓉 , 葛四明 , 曹晓岚 . 腹部超声在健康体检中的应用分析 [J]. 中国医学创新 , 2017 , 14 ( 22 ): 56 - 58 .
SUN R , GE S M , CAO X L . Application analysis of abdominal ultrasound in physical examination [J]. Med Innov China , 2017 , 14 ( 22 ): 56 - 58 .
PEI Y Y , WANG G J , CAO H W , et al . A deep-learning pipeline to diagnose pediatric intussusception and assess severity during ultrasound scanning: a multicenter retrospective-prospective study [J]. NPJ Digit Med , 2023 , 6 ( 1 ): 182 . DOI: 10.1038/s41746-023-00930-8 http://dx.doi.org/10.1038/s41746-023-00930-8 .
XU M L , LV P , LI M Y , et al . Medical image denoising by parallel non-local means [J]. Neurocomputing , 2016 , 195 : 117 - 122 . DOI: 10.1016/j.neucom.2015.08.117 http://dx.doi.org/10.1016/j.neucom.2015.08.117 .
CONTRERAS ORTIZ S H , CHIU T , FOX M D . Ultrasound image enhancement: a review [J]. Biomed Signal Process Control , 2012 , 7 ( 5 ): 419 - 428 . DOI: 10.1016/j.bspc. 2012. 02.002 http://dx.doi.org/10.1016/j.bspc.2012.02.002 .
KRUPINSKI E A , BERBAUM K S , CALDWELL R T , et al . Long radiology workdays reduce detection and accommodation accuracy [J]. J Am Coll Radiol , 2010 , 7 ( 9 ): 698 - 704 . DOI: 10.1016/j.jacr.2010.03.004 http://dx.doi.org/10.1016/j.jacr.2010.03.004 .
NAM J G , KIM M , PARK J , et al . Development and validation of a deep learning algorithm detecting 10 common abnormalities on chest radiographs [J]. Eur Respir J , 2021 , 57 ( 5 ): 2003061 . DOI: 10.1183/13993003.03061-2020 http://dx.doi.org/10.1183/13993003.03061-2020 .
YANG Q , WEI J W , HAO X H , et al . Improving B-mode ultrasound diagnostic performance for focal liver lesions using deep learning: a multicentre study [J]. EBioMedicine , 2020 , 56 : 102777 . DOI: 10.1016/j.ebiom.2020.102777 http://dx.doi.org/10.1016/j.ebiom.2020.102777 .
刘丽 , 谭鹰 , 唐春霖 , 等 . 基于机器学习的早期肝肿瘤超声造影诊断模型建立及应用价值 [J]. 陆军军医大学学报 , 2023 , 45 ( 21 ): 2275 - 2283 . DOI: 10.16016/j.2097-0927.202307131 http://dx.doi.org/10.16016/j.2097-0927.202307131 .
LIU L , TAN Y , TANG C L , et al . Establishment and application value of a diagnostic model for early hepatic tomors using contrast-enhanced ultrasound based on machine learning [J]. J Army Med Univ , 2023 , 45 ( 21 ): 2275 - 2283 . DOI: 10.16016/j.2097-0927.202307131 http://dx.doi.org/10.16016/j.2097-0927.202307131 .
GUPTA P , BASU S M , RANA P , et al . Deep-learning enabled ultrasound based detection of gallbladder cancer in northern India: a prospective diagnostic study [J]. Lancet Reg Health Southeast Asia , 2024 , 24 : 100279 . DOI: 10.1016/j.lansea.2023.100279 http://dx.doi.org/10.1016/j.lansea.2023.100279 .
SALAM A , NAZNINE M , CHOWDHURY M E H , et al . Hybrid neural networks for precise hydronephrosis classification using deep learning [J]. Urology , 2025 , 206 : 17 - 24 . DOI: 10.1016/j.urology.2025.08.005 http://dx.doi.org/10.1016/j.urology.2025.08.005 .
VIANNA P , CALCE S I , BOUSTROS P , et al . Comparison of radiologists and deep learning for US grading of hepatic steatosis [J]. Radiology , 2023 , 309 ( 1 ): e230659 . DOI: 10.1148/radiol.230659 http://dx.doi.org/10.1148/radiol.230659 .
YANG Y , LIU J , SUN C X , et al . Nonalcoholic fatty liver disease (NAFLD) detection and deep learning in a Chinese community-based population [J]. Eur Radiol , 2023 , 33 ( 8 ): 5894 - 5906 . DOI: 10.1007/s00330-023-09515-1 http://dx.doi.org/10.1007/s00330-023-09515-1 .
SCHMAUCH B , HERENT P , JEHANNO P , et al . Diagnosis of focal liver lesions from ultrasound using deep learning [J]. Diagn Interv Imaging , 2019 , 100 ( 4 ): 227 - 233 . DOI: 10.1016/j.diii.2019.02.009 http://dx.doi.org/10.1016/j.diii.2019.02.009 .
DING W Z , MENG Y Q , MA J , et al . Contrast-enhanced ultrasound-based AI model for multi-classification of focal liver lesions [J]. J Hepatol , 2025 , 83 ( 2 ): 426 - 439 . DOI: 10.1016/j.jhep.2025.01.011 http://dx.doi.org/10.1016/j.jhep.2025.01.011 .
NISHIDA N , YAMAKAWA M , SHIINA T , et al . Artificial intelligence (AI) models for the ultrasonographic diagnosis of liver tumors and comparison of diagnostic accuracies between AI and human experts [J]. J Gastroenterol , 2022 , 57 ( 4 ): 309 - 321 . DOI: 10.1007/s00535-022-01849-9 http://dx.doi.org/10.1007/s00535-022-01849-9 .
INTHARAH T , WIRATCHAWA K , WANNA Y , et al . BiTNet: Hybrid deep convolutional model for ultrasound image analysis of human biliary tract and its applications [J]. Artif Intell Med , 2023 , 139 : 102539 . DOI: 10.1016/j.artmed.2023.102539 http://dx.doi.org/10.1016/j.artmed.2023.102539 .
BASU S , GUPTA M , RANA P , et al . RadFormer: Transformers with global-local attention for interpretable and accurate Gallbladder Cancer detection [J]. Med Image Anal , 2023 , 83 : 102676 . DOI: 10.1016/j.media. 2022. 102676 http://dx.doi.org/10.1016/j.media.2022.102676 .
ZHAO W T , ZHANG Z , WANG Z , et al . ECLNet: Center localization of eye structures based on Adaptive Gaussian Ellipse Heatmap [J]. Comput Biol Med , 2023 , 153 : 106485 . DOI: 10.1016/j.compbiomed.2022.106485 http://dx.doi.org/10.1016/j.compbiomed.2022.106485 .
LIANG W X , TADESSE G A , HO D , et al . Advances, challenges and opportunities in creating data for trustworthy AI [J]. Nat Mach Intell , 2022 , 4 ( 8 ): 669 - 677 . DOI: 10.1038/s42256-022-00516-1 http://dx.doi.org/10.1038/s42256-022-00516-1 .
王生生 , 路淑贞 , 曹斌 . 面向隐私保护联邦学习的医学影像目标检测算法 [J]. 计算机辅助设计与图形学学报 , 2021 , 33 ( 10 ): 1553 - 1562 . DOI: 10.3724/SP.J.1089.2021.18416 http://dx.doi.org/10.3724/SP.J.1089.2021.18416 .
WANG S S , LU S Z , CAO B . Medical image object detection algorithm for privacy-preserving federated learning [J]. J Comput Aided Des Comput Graph , 2021 , 33 ( 10 ): 1553 - 1562 . DOI: 10.3724/SP.J.1089.2021.18416 http://dx.doi.org/10.3724/SP.J.1089.2021.18416 .
REN S Q , HE K M , GIRSHICK R , et al . Faster R-CNN: towards real-time object detection with region proposal networks [J]. IEEE Trans Pattern Anal Mach Intell , 2017 , 39 ( 6 ): 1137 - 1149 . DOI: 10.1109/tpami.2016.2577031 http://dx.doi.org/10.1109/tpami.2016.2577031 .
ZHAO Y , LYU W , XU S , et al . DETRs Beat YOLOs on Real-time Object Detection [EB/OL]// 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 2024 : 16965 - 16974 . ( 2024-06-17 ) [ 2026-01-2 9 11: 35 ]. DOI: 10.1109/CVPR52733.2024.01605. Available from: http://dx.doi.org/10.1109/CVPR52733.2024.01605.Availablefrom: https://openaccess.thecvf.com/content/CVPR2024/html/Zhao_DETRs_Beat_YOLOs_on_Real-Time_Object_Detection_CVPR_2024_paper.html https://openaccess.thecvf.com/content/CVPR2024/html/Zhao_DETRs_Beat_YOLOs_on_Real-Time_Object_Detection_CVPR_2024_paper.html
JIA H J , ZHANG L J , LIANG X M , et al . DPDB-YOLO: a lightweight YOLOv13 cherry tomato ripeness detection method with adaptive extraction module and multi-scale feature fusion architecture [J]. Ind Crops Prod , 2025 , 238 : 122419 . DOI: 10.1016/j.indcrop.2025.122419 http://dx.doi.org/10.1016/j.indcrop.2025.122419 .
KARAKO K , MIHARA Y , ARITA J , et al . Automated liver tumor detection in abdominal ultrasonography with a modified faster region-based convolutional neural networks (Faster R-CNN) architecture [J]. Hepatobiliary Surg Nutr , 2022 , 11 ( 5 ): 675 - 683 . DOI: 10.21037/hbsn-21-43 http://dx.doi.org/10.21037/hbsn-21-43 .
DADOUN H , ROUSSEAU A L , DE KERVILER E , et al . Deep learning for the detection, localization, and characterization of focal liver lesions on abdominal US images [J]. Radiol Artif Intell , 2022 , 4 ( 3 ): e210110 . DOI: 10.1148/ryai.210110 http://dx.doi.org/10.1148/ryai.210110 .
WU C H , SHEU J C , CHOU P L , et al . Automatic real-time detection and diagnosis of liver tumor with ultrasound [J]. J Hepatocell Carcinoma , 2025 , 12 : 1599 - 1611 . DOI: 10.2147/JHC.S524311 http://dx.doi.org/10.2147/JHC.S524311 .
ZHAO Z J , CHEN Z R , VOROS S , et al . Real-time tracking of surgical instruments based on spatio-temporal context and deep learning [J]. Comput Assist Surg , 2019 , 24 ( sup1 ): 20 - 29 . DOI: 10.1080/24699322.2018.1560097 http://dx.doi.org/10.1080/24699322.2018.1560097 .
0
浏览量
0
下载量
0
CSCD
关联资源
相关文章
相关作者
相关机构
京公网安备11010802024621