Level II Ultrasound Standard Plane Recognition for Mid Pregnancy by a Specific ResNet20
Xiong Runqing1&, Cai Jiaxin2&*, Zheng Liping1, Ma Duo1, Dai Chenquan3
1(Department of Ultrasound, The Second Affiliated Hospital of Xiamen Medical College, Xiamen 361000, Fujian, China) 2(School of Mathematics and Statistics, Xiamen University of Technology, Xiamen 361000, Fujian, China) 3(School of Computer and Information Engineering, Xiamen University of Technology, Xiamen 361000, Fujian, China)
Abstract：To address the problems of experience-dependency, low efficiency and wrong diagnosis risks in standard section recognition during second-trimester fetal level II prenatal ultrasound screening, this study explored the application of convolutional neural network (CNN) technology for automatic recognition of multi-section ultrasound images, aiming to improve screening accuracy and real-time performance. Ultrasound images of second-trimester fetuses from 200 pregnant women were collected using GE E8 and Samsung WS80A color Doppler devices, covering 10 standard fetal sections (1 869 images in total). All images were normalized to 227 pixels×227 pixels. A specially designed ResNet20 model was constructed, featuring a seven-layer architecture with bottleneck residual modules, and the softmax function was used for classification probability output. During the training, 90% of the images were used as the training set, and 10% of them as the test set. The optimization employed a multi-class cross-entropy loss with a learning rate of 0.1 and a momentum optimizer. The ResNet20 model achieved an overall recognition accuracy of 88.2%, with 100% precisions in identifying the cerebellar transverse section, bilateral eye transverse section, and spinal sagittal section. In comparison, AlexNet and VGG16 showed accuracies of 80.2% and 73.3%, respectively, with statistically significant differences (P<0.001). The average testing time was 33 seconds per case, meeting clinical real-time requirements. The proposed ResNet20 model enabled efficient automatic recognition of second-trimester fetal level II standard sections, demonstrating good performance and clinical application prospects, contributes to the intelligent and standardized improvement of ultrasound screening. The code is open-source at https://github.com/Chenan7/Resnet-19.
熊润青, 蔡加欣, 郑丽萍, 马多, 戴陈泉. 基于特制ResNet20的中孕期Ⅱ级产前超声检查胎儿切面识别[J]. 中国生物医学工程学报, 2025, 44(5): 533-540.
Xiong Runqing, Cai Jiaxin, Zheng Liping, Ma Duo, Dai Chenquan. Level II Ultrasound Standard Plane Recognition for Mid Pregnancy by a Specific ResNet20. Chinese Journal of Biomedical Engineering, 2025, 44(5): 533-540.
[1] 李胜利,陈秀兰,姚远,等.产前超声诊断进修医师培训质量控制[J/CD].中华医学超声杂志:电子版, 2009,6(4):630-637. [2] Chen Hao, Ni Dong, Qin Jing, et al.Standard plane localization in fetal ultrasound via domain transferred deep neural networks[J].IEEE J Biomed Health Inform, 2015,19(5): 1627-1636. [3] Wu Lingyun, Cheng Jiezhi, Li Shengli, et al.FUIQA: fetal ultrasound image quality assessment with deep convolutional networks[J].IEEE Transactions on Cybernetics, 2017,47(5):1336-1349. [4] Potka S, Wodarczyk T, KlasaA, et al.FetalNet: Multi-task deep learning framework for fetal ultrasound biometric measurements[C]//International Conference on Neural Information Processing, Indonesia: Springer, 2021: 257-265. [5] Heuvel T, Petros H, Santini S, et al.Automated fetal head detection and circumference estimation from free-hand ultrasound sweeps using deep learning in resource-limited countries[J].Ultrasound Med Biol, 2019, 45: 773-785. [6] Wang Xiaoli, Liu Zhonghua, Du Yongzhao, et al.Recognition of Fetal Facial Ultrasound Standard Plane Based on Texture Feature Fusion[J].Comput Math Methods Med, 2021, 6656942. [7] Jang J, Park Y, Kim B, et al.Automatic estimation of fetal abdominal circumference from ultrasound images[J].IEEE J Biomed Health Inform,2018, 22(5):1512-1520. [8] 中华医学会超声医学分会妇产超声学组, 国家卫生健康委妇幼司全国产前诊断专家组医学影像组.超声产前筛查指南[J].中华超声影像学杂志, 2022, 31(1):1-12. [9] 周涛,刘赟璨,陆惠玲,等.ResNet及其在医学图像处理领域的应用:研究进展与挑战[J].电子与信息学报,2022,44(1):149-167. [10] Salim I, Cavallaro A, Ciofoloveit C, et al.Evaluation of automated tool for two-dimensional fetal biometry[J].Ultrasound in Obstetrics & Gynecology.2019;54(5):650-654. [11] Espinoza J, Good S, Russell E, et al.Does the use of automated fetal biometry improve clinical work flow efficiency?[J].Journal of Ultrasound in Medicine, 2013, 32(5):847-850. [12] Yaqub M, Kelly B, Papageorghiou AT, et al.A deep learning solution for automatic fetal neurosonographic diagnostic plane verification using clinical standard constraints.[J].Ultrasound in Medicine & Biology, 2017,43(12):2925-2933. [13] 余镇, 吴凌云, 倪东,等.基于深度学习的胎儿颜面部超声标准切面自动识别[J].中国生物医学工程学报, 2017, 36(3):9: 267-275. [14] Arnaout R, Curran L, Zhao Yingli, et al.An ensemble of neural networks provides expert-level prenatal detection of complex congenital heart disease[J].Nature Medicine, 2021, 27(5):882-891. [15] Coronado-Gutiérrez D, Eixarch E, Monterde E, et al.Automatic deep learning-based pipeline for automatic delineation and measurement of fetal brain structures in routine mid-trimester ultrasound images[J].Fetal Diagn Ther.2023, 50(6):480-490. [16] Ondeck CL, Pretorius D, McCaulley J, et al.Ultrasonographic prenatal imaging of fetal ocular and orbital abnormalities[J].Surv Ophthalmol.2018, 63(6):745-753. [17] 郭家菊,秦越,李胜利.胎儿先天性白内障产前超声诊断及研究分析[J].中华医学超声杂志(电子版),2021,18(3):285-289. [18] Pertl B, Eder S, Stern C, et al.The fetal posterior fossa on prenatal ultrasound imaging: normal longitudinal development and posterior fossa anomalies[J].Ultraschall Med, 2019, 40(6):692-721. [19] Schlemper J, Oktay O, Schaap M, et al.Attention gated networks: Learning to leverage salient regions in medical images[J].Med Image Anal 2019,53: 197-207. [20] 刘中华,王小莉,吕国荣,等.人工智能自动识别胎儿颜面部超声标准切面的研究[J].中国医学物理学杂志, 2021, 38(12):1575-1578.
