Nuclear segmentation in four-dimensional label-free microscopy images for predicting live birth potential of mouse embryos

Assisted reproductive technology (ART), including in vitro fertilization, is one of the most common treatments for infertility. Assessment of embryo quality is an important step in selecting embryos with high birth potential in ART. However, it currently relies on visual assessment by experts, and the birth rate remains low. We previously developed a deep learning method to predict the birth of mouse embryos by quantifying the morphological features of cell nuclei. This method involves cell nuclear segmentation on fluorescence microscopy images, but fluorescence labeling of nuclei is not feasible in medical applications. Here, we developed FL 2 -Net, a nuclear segmentation method for time-series three-dimensional bright-field microscopy images of mouse embryos without fluorescence labeling. FL 2 -Net is a deep learning-based segmentation method exploiting the spatiotemporal features of preimplantation development. We showed that FL 2 -Net outperformed existing state-of-the-art segmentation methods. In addition, we performed the birth prediction of mouse embryos from the nuclear features quantified by bright-field microscopy image segmentation. The birth prediction accuracy using FL 2 -Net (81.63%) exceeded those using other methods and those of experts’ visual assessment (55.32%). We expect that FL 2 -Net, which can quantify nuclear features of embryos non-invasively and with high accuracy, might be useful in ART. The code is publicly available at https://github.com/funalab/FL2-Net . • FL 2 -Net is a segmentation method for 3D bright-field microscopy images. • FL 2 -Net outperformed state-of-the-art segmentation methods. • It enables non-invasive extraction of spatiotemporal features during embryogenesis. • We performed birth prediction of mouse embryos based on the quantified features. • The prediction achieved greater accuracy than that of experts’ visual assessment.