Label-free three-dimensional subcellular phenotyping of mouse embryos by holotomography enables early prediction of blastocyst formation

Abstract Imaging-based assessment of embryo competence is central to embryo selection in in vitro fertilization (IVF), yet current practice relies largely on gross morphology with limited access to three-dimensional (3D) subcellular information. Here, we present a label-free framework for quantitative 3D embryo phenotyping using holotomography (HT). Time-lapse HT enables volumetric imaging of mouse embryos from the 2-cell stage to the blastocyst with subcellular contrast. Early-stage quantitative analysis shows that embryos that subsequently form blastocysts exhibit higher blastomere counts, lower blastomere heterogeneity, and tighter nuclear packing, whereas embryos that arrest display enlarged blastomeres, elevated cytoplasmic refractive-index heterogeneity, and fewer, larger nuclei. Machine-learning models trained on these features demonstrate the feasibility of predicting blastocyst formation (AUC up to 0.963). These findings establish HT as a promising label-free approach for quantitative 3D phenotyping of early embryos and for identifying predictive subcellular features associated with developmental outcome.