Embryo development stage prediction algorithm for automated time lapse incubators.

Area of Science:

• Embryology
• Reproductive Medicine
• Medical Imaging Analysis

Background:

• Time-lapse microscopy is crucial for monitoring embryo development and selecting viable embryos for fertilization.
• Current embryo selection processes are time-consuming and resource-intensive, impacting pregnancy success rates.
• Automated tools are needed to improve embryo quality evaluation and predict developmental stages.

Purpose of the Study:

• To develop an automated method for predicting embryo development stages using time-lapse microscopy images.
• To enhance the accuracy of embryo selection for in vitro fertilization (IVF).

Main Methods:

• A two-classifier vote-based method was employed for embryo image classification.
• Features were extracted using a Convolutional Neural Network (CNN).
• Embryo development stage prediction was achieved by comparing the confidence scores of two classifiers.

Main Results:

• The system achieved an overall accuracy of 97.62% on a test set.
• Combining CNN with a Discriminant classifier yielded the most effective results.
• The proposed method improved the accuracy of CNN-based classification by 1.04%.

Conclusions:

• An automated approach for predicting embryo development stages from time-lapse microscopy images was successfully developed.
• The method utilizes high-complexity image features extracted by CNN and a novel two-classifier comparison strategy.
• This advancement offers practical implications for improving embryo selection in IVF procedures.