Scholarlink

<p>Accurate segmentation of cervix in intrapartum ultrasound images and measurement of cervical dilation (CD) are crucial for monitoring labor progress and predicting delivery outcomes. However, the poor quality of ultrasound images like artifacts and missing boundaries makes the measurement inaccurate and time- and effort-consuming. This paper proposes CDNet, a network for automatic measurement of CD. First, CDNet accurately segments the cervix mask, and then the center of the mask is located to find the horizontal and vertical intersection points to calculate the transverse diameter and anteroposterior diameter of CD. A shape-constraint loss function is used to improve segmentation accuracy by the prior convex shape. Experiments show that CDNet achieved the lowest Relative Volume Error of 15.56 ± 1.80%, the highest Dice of 89.55 ± 0.47%, and the lowest 95% Hausdorff distance of 11.48 ± 0.63&#xa0;mm. The measurement of transverse diameter and anteroposterior diameter based on CDNet achieved a mean absolute error of 1.79 ± 0.24&#xa0;mm and 2.67 ± 0.22&#xa0;mm respectively. Moreover, CDNet also outperformed its counterparts in further experiments on two additional fetal head datasets. In conclusion, our method can achieve automatic CD measurement with good performance and may help assess labor progress in the future.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An automatic measurement of cervix dilation in intrapartum ultrasound image

  • 期刊论文

  • 发表日期：

    2025年1月17日

查看全文
  • Yifan Zhang,
  • Jinwen Zhu,
  • Shun Long,
  • Jieyun Bai,
  • Yaosheng Lu,
  • Gaowen Chen

摘要

Accurate segmentation of cervix in intrapartum ultrasound images and measurement of cervical dilation (CD) are crucial for monitoring labor progress and predicting delivery outcomes. However, the poor quality of ultrasound images like artifacts and missing boundaries makes the measurement inaccurate and time- and effort-consuming. This paper proposes CDNet, a network for automatic measurement of CD. First, CDNet accurately segments the cervix mask, and then the center of the mask is located to find the horizontal and vertical intersection points to calculate the transverse diameter and anteroposterior diameter of CD. A shape-constraint loss function is used to improve segmentation accuracy by the prior convex shape. Experiments show that CDNet achieved the lowest Relative Volume Error of 15.56 ± 1.80%, the highest Dice of 89.55 ± 0.47%, and the lowest 95% Hausdorff distance of 11.48 ± 0.63 mm. The measurement of transverse diameter and anteroposterior diameter based on CDNet achieved a mean absolute error of 1.79 ± 0.24 mm and 2.67 ± 0.22 mm respectively. Moreover, CDNet also outperformed its counterparts in further experiments on two additional fetal head datasets. In conclusion, our method can achieve automatic CD measurement with good performance and may help assess labor progress in the future.