Embryo Viability Scoring With 3D Neural Networks From Time-Lapse Video

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current embryo grading systems in IVF rely heavily on subjective human judgment, leading to high inter- and intra-embryologist variability, inefficiency, and lack of reproducibility, making it difficult to select the embryo with the highest pregnancy potential for single embryo transfer.

Innovation Solution

A deep learning system utilizing a 3D convolutional neural network (CNN) analyzes time-lapse embryo videos to provide an objective viability score, capturing both spatial and temporal features without manual annotation, enabling automatic embryo ranking and selection.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If manual embryo grading by embryologists is used, then embryo quality assessment can be performed, but high inter- and intra-embryologist variability and subjective judgment reduce reliability and reproducibility

Engineering Contradiction:
Improvereproducibility of embryo viability assessmentVSAvoidcomplexity of grading system
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent replaces the manual mechanical grading process performed by embryologists with an automated deep learning system using 3D convolutional neural networks. The system processes time-lapse video data of embryos and generates viability scores without human intervention, eliminating subjective variability while maintaining assessment capability. This substitution of mechanical human judgment with automated computational analysis directly resolves the contradiction between reliability and complexity.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Reliability

If multiple embryos are implanted to increase success rate, then pregnancy potential increases, but probability of multiple pregnancies and antenatal complications increases

Engineering Contradiction:
Improvepregnancy success rateVSAvoidrisk of multiple pregnancies and complications
Core Design Contradiction:
ReliabilityVSObject-affected harmful factors

Solution Approach 1:

The patent changes the parameter of embryo selection from subjective categorical grading to continuous viability scoring based on deep learning analysis. By transforming the assessment parameter into a precise numerical score that predicts pregnancy potential, the system enables confident selection of a single high-potential embryo, allowing single embryo transfer while maintaining high success rates and eliminating the risks associated with multiple pregnancies.

Inventive Principle:
Principle #35Parameter changes

3Measurement precision

If manual annotation of each embryo image is performed, then detailed morphological features can be analyzed, but the process becomes time-consuming and labour-intensive requiring up to 1 hour per patient

Engineering Contradiction:
Improveprecision of morphological feature analysisVSAvoidthroughput of embryo assessment
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent replaces manual annotation with automated deep learning-based image analysis. The 3D convolutional neural network automatically extracts and analyzes morphological features from time-lapse video data without requiring human annotation. This substitution maintains measurement precision through sophisticated feature extraction while increasing productivity by processing multiple embryos simultaneously without the time constraints of manual analysis.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

4Measurement precision

If current grading systems analyze only 2 to 4 isolated time frames, then specific developmental checkpoints are evaluated, but comprehensive temporal development patterns are missed

Engineering Contradiction:
Improveaccuracy of developmental checkpoint assessmentVSAvoidloss of temporal development information
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

The patent adds the temporal dimension to embryo assessment by analyzing complete time-lapse video sequences rather than isolated frames. The 3D convolutional neural network processes videos containing multiple time points, capturing the dynamic development patterns and temporal progression of embryonic growth. This dimensional expansion from static frame analysis to dynamic video analysis preserves comprehensive temporal information while maintaining precision in developmental checkpoint assessment.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Data Source

PatentEP3723640B1Systems and methods for estimating embryo viability
Publication Date: 2025.11.05 VITROLIFE AS
  • EP3723640B1 patent drawingFigure 1
  • EP3723640B1 patent drawingFigure 2
  • EP3723640B1 patent drawingFigure 3A~3B

AI summary

A computer-implemented method, including the steps of: receiving video data of a human embryo, the video data representing a sequence of images of the human embryo in chronological order; applying at least one three-dimensional (3D) artificial neural network (ANN) to the video data to determine a viability score for the human embryo, wherein the viability score represents a likelihood that the human embryo will result in a viable embryo or a viable fetus; and outputting the viability score.