Multimodal Learning for Embryo Viability Prediction in Clinical IVF

TL;DR

This paper presents a multimodal AI model that combines time-lapse videos and electronic health records to improve and automate embryo viability prediction in IVF, aiming to enhance success rates and reduce manual assessment variability.

Contribution

The study introduces a novel multimodal model integrating video and EHR data for embryo viability prediction, addressing modality differences and enhancing clinical IVF decision-making.

Findings

Effective combination of video and EHR data improves prediction accuracy.

Multimodal approach reduces manual assessment time and variability.

Model demonstrates potential for scalable, automated embryo viability assessment.

Abstract

In clinical In-Vitro Fertilization (IVF), identifying the most viable embryo for transfer is important to increasing the likelihood of a successful pregnancy. Traditionally, this process involves embryologists manually assessing embryos' static morphological features at specific intervals using light microscopy. This manual evaluation is not only time-intensive and costly, due to the need for expert analysis, but also inherently subjective, leading to variability in the selection process. To address these challenges, we develop a multimodal model that leverages both time-lapse video data and Electronic Health Records (EHRs) to predict embryo viability. One of the primary challenges of our research is to effectively combine time-lapse video and EHR data, owing to their inherent differences in modality. We comprehensively analyze our multimodal model with various modality inputs and integration approaches. Our approach will enable fast and automated embryo viability predictions in scale for clinical IVF.