基于深度学习的胎儿头部和腹部超声影像智能测量技术研究  

Research on Intelligent Measurement Technology of Fetal Head and Abdomen Ultrasound Image Based on Deep Learning

作　　者：夏书剑[1] 孙艳秋[1] 张柯欣[1] XIA Shujian;SUN Yanqiu;ZHANG Kexin(Liaoning University of Traditional Chinese Medicine,Shenyang 110847,Liaoning,China)

机构地区：[1]辽宁中医药大学,辽宁沈阳110847

出　　处：《中华中医药学刊》2023年第7期98-103,I0021-I0023,共9页Chinese Archives of Traditional Chinese Medicine

基　　金：国家自然科学基金项目(82274580);国家级大学生创新创业项目(202210162019);辽宁省教育厅科学技术研究重点攻关项目(L2020059);辽宁省教育厅科学研究项目(LJKZ0894);辽宁省普通高等教育本科教学改革研究项目(2021-549);省级大学生创新创业项目(S202210162003X);辽宁中医药大学自然科学类重点项目(2021018);辽宁中医药大学人文社会科学研究项目(2021004);辽宁中医药大学人文社会科学一般研究项目(2019-lnzy001);辽宁中医药大学教学改革研究项目(LNZYJG2022063,LNZYJG2020135)。

摘　　要：超声影像技术具有无创、无辐射、廉价和实时的优点,其被广泛应用于产前检查诊断中。在超声影像检查中胎儿头围、腹围以及股骨长等参数的测量一直是评估胎儿发育状况的重要依据。但目前超声图像质量易受噪声、阴影和混响的影响,同时临床医生的经验技术以及被栓者状态也会使得医生手动测量的胎儿生物参数结果不够准确。为了避免外界干扰,减轻医生的工作负担,使用人工智能技术辅助胎儿超声诊断显得尤为重要。目前在医学图像分割网络中,“编码-解码”U型结构网络(Unet)因其架构简单,精度高,易实现的优势而被广泛使用。鉴于此,针对胎儿头部腹部超声影像,基于Unet语义分割网络模型,针对胎儿超声图像头围特征测量的轮廓变化较大、图像干扰较多等问题,首先实现超声图像特征提取,然后对超声图像中的头骨和腹围进行分割,将分割后区域边界根据最小二乘法拟合椭圆等过程实现头围、腹围超声图像的自动测量。同时构建了产前超声智能测量系统,实现了超声图像中胎儿头围、腹围等生物参数的自动测量功能,具有重要临床意义。最后对基于深度学习的图像智能处理技术在中医领域的应用情况、面临问题和应用前景进行探讨。Ultrasound imaging technology has the advantages of non-invasive,non-radiation,low cost and real-time and is widely used in prenatal diagnosis.The measurement of fetal head circumference,abdominal circumference and femur length in ultrasound imaging is always an important basis for evaluating fetal development.However,the quality of ultrasound images is susceptible to noise,shadow and reverberation.At the same time,the experience and skills of clinical doctors,as well as the patients'status,can also make the results of manually measured fetal biological parameters inaccurate.In order to avoid external interference and reduce the workload of doctors,it is particularly important to use artificial intelligence technology to assist fetal ultrasound diagnosis.In the current medical image segmentation network,Unet('encoding-decoding'U-shaped structure network)is also widely used because of its simple architecture,high precision and easy implementation.In view of this,the article focused on fetal head and abdomen ultrasound images.Based on the Unet semantic segmentation network model,in order to address the issues of significant contour changes and high image interference in fetal ultrasound image head circumference feature measurement,we first achieved ultrasound image feature extraction,and then segmented the head and abdominal circumference in the ultrasound image.After segmentation,the boundary of the segmented area was fitted with ellipses using the least squares method to achieve head circumference and abdominal circumference automatic measurement of ultrasound images.At the same time,an intelligent prenatal ultrasound measurement system was constructed to realize the automatic measurement of fetal head circumference,abdominal circumference and other biological parameters in ultrasound images,which had important clinical significance.Finally,the application,problems and application prospects of image intelligent processing technology based on deep learning in the field of traditional Chinese medicine were discusse

关 键 词：卷积神经网络 深度学习 智能测量 超声影像 

分 类 号：R445.1[医药卫生—影像医学与核医学] R2-03[医药卫生—诊断学]