Automated monitoring of human embryonic cells up to the 5-cell stage in time-lapse microscopy images

Abstract

Measurement of the proliferative behavior of human embryonic cells in vitro is important to many biomedical applications ranging from basic biology research to advanced applications, such as determining embryo viability during in vitro fertilization (IVF) treatments. Automated prediction of the embryo viability, by tracking cell divisions up to the 4-cell stage, improves embryo selection and may lead to increased success rates in IVF pregnancies. Recent research in cell biology has suggested that tracking cell divisions beyond the 4-cell stage further improves embryo selection. In the current state-of-the-art, later events (e.g., time to reach the 5-cell stage) can only be assessed manually. In this work we automatically predict the number of cells at every time point, and predict when the embryo divides beyond four cells in a time-lapse microscopy sequence. Our approach employs a conditional random field (CRF) that compactly encodes various aspects of the evolving embryo and estimates the number of cells at each time step via exact inference. We demonstrate the effectiveness of our method on a data set of 33 developing human embryos