谷歌浏览器插件
订阅小程序
在清言上使用

CYTOPLASMIC MORPHOLOGICAL CHARACTERISTICSAFFECT 2PN DETECTION IN AN AUTOMATIC PRONUCLEAR NUMBER DETECTION SYSTEM USING DEEP LEARNING TECHNOLOGY.

Fertility and sterility(2021)

引用 0|浏览8
暂无评分
摘要
We have been developing an automatic pronuclear (PN) number detection system using AI in order to reduce implantation failures caused by misjudgment of fertilization result with observation by clinical embryologists.Since our algorithm outputs the number of PN based on detected contour of the pronucleus, morphological characteristics in the cytoplasm may reduce the accuracy by interfering contour detection and misevaluating detected contour.Therefore, we investigated, with the aim of improving the accuracy of PN number detection, whether morphological characteristics in the cytoplasm affect the accuracy of pronuclear number detection. After ICSI, we targeted 4947 oocytes evaluated as 2PN by clinical embryologists, the PN number of each embryo was computed by the pronucleus number automatic detection system.(version 2019)Embryos having morphological characteristics in the cytoplasm were designated as CLCG(centrally located cytoplasmic granularity) group, Vesicle group, halo group respectively and embryos without the above characteristics were designated as controls.The respective rates were then compared.The correct 2PN answer rates for these groups were statistically processed using the χ2 test and multiple comparisons by the Ryan method to indicate the affect by cytoplasmic morphologic characteristics. The correct 2PN answer rate was 93.6% (2565/2739) in the control group, 86.9% (764/879) in the CLCG group, 89.9% (664/739) in the Vesicle group, and 95.4% (563/590) in the halo group.The correct answer rate was significantly lower in the CLCG group and Vesicle group than in the Control group.In addition, the correct answers rate was significantly lower in the CLCG group and Vesicle group than in the halo group. We have determined that CLCG and Vesicle characteristics affect the accuracy of PN number detection. On the other hand, halo characteristic does not affect the accuracy of PN number detection. CLCG and Vesicle appear in the center of the cytoplasm and overlap with the pronucleus to inhibit the contour detection of the pronucleus, but halo is found at the margin of the cytoplasm and is considered not to interfere with detection of the pronucleus. We are continuing to work on improving the algorithm and the accuracy of PN detection of embryos with morphological characteristics.
更多
查看译文
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要