Work

EmbryoSense: An Analytics and AI Platform for Time-Lapse Embryo Assessment and Decision Support

Year: 2026

Type: article

Abstract: Time-lapse imaging (TLI) has enabled continuous observation of embryo devel-opment, yet clinical decision-making can remain variable and difficult to bench-mark against a clinic’s own historical outcomes. EmbryoSense is an integrated analytics and AI/ML platform designed to support embryo assessment through (i) automated, customizable reporting and cohort-based comparisons and (ii) pre-dictive models that estimate developmental and quality outcomes from TLI-derived data. The Analytics module generates patient- and clinic-oriented reports and enables “smart” comparisons of current attempts against targeted historical subgroups defined by key variables (e.g., maternal age, embryologist score, infer-tility cause) and morphokinetic parameters (e.g., tPNf, t2, t6–t5). The AI/ML module provides prediction capabilities including Day5 quality prediction from kinetic data up to Day3, blastocyst-stage prediction using TLI data up to 84 hours, and dynamic time grading, supported by continuous monitoring of embryo dynamics via high-order image-extracted features (e.g., embryo area, solidity, speed, eccentricity, inertia tensors). In the reported evaluation, EmbryoSense models achieved AUC 0.87 for Day3 grade (Bad/Good), AUC 0.846 for blasto-cyst-stage prediction (Day3.5), and AUC 0.77 for Day5 grade (Bad/Good), indi-cating strong discriminative performance for early decision support.

Source: Zenodo (CERN European Organization for Nuclear Research)

DOI: 10.5281/zenodo.18447691