EmbryosFormer: Deformable Transformer and Collaborative Encoding-Decoding for Embryos Stage Development Classification

TL;DR

EmbryosFormer is a novel deformable transformer-based model that automatically detects and classifies embryo cell divisions from time-lapse images, improving accuracy and efficiency in IVF embryo viability assessment.

Contribution

It introduces a collaborative encoder-decoder deformable transformer architecture for embryo stage classification, integrating temporal coherence and multi-head learning.

Findings

Achieved state-of-the-art accuracy on mouse embryo dataset.

Demonstrated effective temporal modeling for embryo development stages.

Reduced manual effort in embryo viability assessment.

Abstract

The timing of cell divisions in early embryos during the In-Vitro Fertilization (IVF) process is a key predictor of embryo viability. However, observing cell divisions in Time-Lapse Monitoring (TLM) is a time-consuming process and highly depends on experts. In this paper, we propose EmbryosFormer, a computational model to automatically detect and classify cell divisions from original time-lapse images. Our proposed network is designed as an encoder-decoder deformable transformer with collaborative heads. The transformer contracting path predicts per-image labels and is optimized by a classification head. The transformer expanding path models the temporal coherency between embryo images to ensure monotonic non-decreasing constraint and is optimized by a segmentation head. Both contracting and expanding paths are synergetically learned by a collaboration head. We have benchmarked our proposed EmbryosFormer on two datasets: a public dataset with mouse embryos with 8-cell stage and an in-house dataset with human embryos with 4-cell stage. Source code: https://github.com/UARK-AICV/Embryos.