AI Embryo Ploidy Prediction for Non-Invasive IVF Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for embryo selection in IVF suffer from intra- and inter-observer variability in morphological assessment and invasiveness of PGT-A, lacking a standardized, non-invasive approach for predicting embryo ploidy.
Innovation Solution
A non-invasive method using artificial intelligence to predict embryo ploidy by analyzing static images with machine and deep learning models, incorporating morphological and clinical features, and generating a ploidy status prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If morphological assessment and morphokinetic analysis are used for embryo selection, then embryo viability evaluation is performed, but intra- and inter-observer variability occurs due to unstandardized methods
Solution Approach 1:
The patent replaces manual morphological assessment and morphokinetic analysis with an automated deep learning system that processes time-lapse microscopy images. The system uses convolutional neural networks to objectively evaluate embryo morphology and predict ploidy status, eliminating human observer variability while maintaining high accuracy in embryo viability evaluation.
Solution Approach 2:
The patent creates a digital replica of the embryo assessment process through trained deep learning models. These models learn from labeled training data containing morphological features and morphokinetic parameters, then apply this learned knowledge to new embryos, ensuring consistent and reproducible evaluations across different users and time points.
2Measurement precision
If PGT-A (pre-implantation genetic testing for aneuploidy) is used for embryo selection, then ploidy status is determined, but the method is invasive and costly
Solution Approach 1:
The patent introduces a non-invasive deep learning-based prediction system as an intermediary between traditional morphological assessment and invasive PGT-A. The system uses time-lapse microscopy images and morphological features to predict ploidy status, providing a screening tool that identifies embryos likely to be euploid or aneuploid without requiring tissue biopsy, thereby reducing or eliminating the need for invasive PGT-A procedures.
Solution Approach 2:
The patent replaces the invasive mechanical process of trophectoderm biopsy with a non-invasive computational analysis system. The deep learning model analyzes external observable features from time-lapse images to infer internal genetic status, substituting physical tissue sampling with information processing that causes no harm to the embryo.
3Ease of operation
If deep learning is used to predict embryo quality, then morphological assessment limitations are alleviated, but standardization for ploidy prediction remains insufficient
Solution Approach 1:
The patent performs preliminary training of deep learning models using large datasets of labeled embryo images with known ploidy outcomes. This pre-training phase establishes standardized prediction criteria and performance benchmarks before clinical application. The models are trained on diverse data including morphological features, morphokinetic parameters, and confirmed ploidy status to learn standardized patterns associated with euploid and aneuploid embryos.
Solution Approach 2:
The patent implements feedback mechanisms where model predictions are compared against actual PGT-A results when available, allowing for continuous validation and refinement of prediction accuracy. The system provides feedback on prediction confidence levels and can be retrained with new data to improve standardization and precision over time, creating a self-improving standardized assessment tool.
Data Source
AI summary
The present disclosure encompasses systems and methods for predicting embryo ploidy. Specific embodiments encompass methods of non-invasively predicting ploidy status of an embryo, by receiving a dataset with a static image of the embryo, analyzing the static image by one or more machine and/or deep learning model via one or more classification task applied to the dataset; and generating an output prediction of the ploidy status of the embryo. Particular methods relate to methods wherein the dataset additionally includes one or more clinical and/or morphological features for the embryo. Embodiments also relate to predicting embryo viability and/or improving embryo selection, such as during in vitro fertilization, and uses thereof.