摘要
为有效克服胎儿头部超声图像中存在的干扰问题,实现对胎儿头围的精确测量,提出一种基于融合边界框高置信度区域信息的超声胎儿头围测量算法。首先,通过U-Net分割网络提取胎儿头部图像感兴趣区域;其次,利用YOLOv3检测网络获取胎儿头部边界框,结合边界框高置信度区域信息,筛选头部感兴趣区域高置信度边缘点;最后,采用直接最小二乘法对高置信度边缘点进行椭圆拟合,计算胎儿头围结果。实验结果表明:该算法可有效克服图像质量的干扰,提高超声胎儿头围测量精度。
In order to effectively overcome the interference in the fetal head image and achieve accurate measurement of the fetal head circumference,this paper proposes an ultrasonic fetal head circumference measurement algorithm that fuse the high confidence region information of the bounding box.First,extract the region of interest(ROI)of the fetal head through the U-Net segmentation network;secondly,use the YOLOv3 detection network to obtain the head bounding box,and combine the high confidence region information of the bounding box to filter the high confidence edge points of the head region of interest;Finally,the direct least square method is used to ellipse fit the high-confidence edge points to calculate the fetal head circumference.Experimental results show that the algorithm can effectively overcome the interference of image quality and improve the measurement accuracy of ultrasound fetal head circumference.
作者
汪金婷
杨丰
陈琪
Wang Jinting;Yang Feng;Chen Qi(School of Biomedical Engineering Southern Medical University,Guangzhou 510515,China)
出处
《自动化与信息工程》
2021年第1期7-11,共5页
Automation & Information Engineering
基金
国家自然科学基金项目(61771233)。
