激光生物学报2024，Vol.33Issue(3)：209-216,8.DOI:10.3969/j.issn.1007-7146.2024.03.003
基于DE-3D U-Net的斑马鱼后脑血管共聚焦显微图像分割和量化测量
Segmentation and Quantification of Zebrafish Hindbrain Vessel Confocal Images Based on DE-3D U-Net
摘要
Abstract
The choroid plexues in the zebrafish hindbrain controls the exchange of substances between brain and its vessels,and the vascular abnormalities around the choroid plexues can lead to associated cerebrovascular diseases.We first performed serial imaging of the major blood vessels in the hindbrain region of zebrafish,and obtained a series of images of zebrafish be-tween 43 hours and 63 hours after fertilization using confocal microscopy to photograph embryos under normal conditions and after cylindrospermopsin treatment.Afterwards,the DE-3D U-Net was developed for the segmentation of the hindbrain images.The segmentation results showed that the class pixel accuracy of DE-3D U-Net could reach to 86.67%,93.18% and 83.74% for the three main vessels and the segmentation results were used to plot vascularization curves.The quantitative results showed that the radii of the three vessels showed a reduction of 18% after treatment with high concentrations of cylindrospermopsin,and the angle of the mesencephalic vein showed a reduction of 30° to 50°.This study proposed a neural network for segmentation of hindbrain vessels of zebrafish,and it provided a basis for research on the development of zebrafish choroid plexsus.关键词
斑马鱼/后脑血管/共聚焦显微成像/3D U-Net神经网络/柱孢藻毒素Key words
zebrafish/hindbrain vessels/confocal microscopy/3D U-Net deep-learning network/cylindrospermopsin分类
信息技术与安全科学引用本文复制引用
郑文虎,李辉,陈冲,王林波..基于DE-3D U-Net的斑马鱼后脑血管共聚焦显微图像分割和量化测量[J].激光生物学报,2024,33(3):209-216,8.基金项目
国家自然科学基金项目(62205368). （62205368）
