Automatic blastomere recognition from a single embryo image

Area of Science:

• Embryology
• Medical Imaging
• Computer Vision

Background:

• Accurate blastomere counting is crucial for human embryo viability assessment.
• Challenges exist in automated blastomere recognition due to embryo transparency and blastomere overlap in single images.

Purpose of the Study:

• To develop and validate an automated method for blastomere recognition and counting from single human embryo images.
• To improve the efficiency and accuracy of embryo viability assessment.

Main Methods:

• Image preprocessing involving edge detection, point filtering, dilation, and erosion to isolate blastomere edges.
• Least square circle fitting algorithm for automatic blastomere recognition and counting.
• Validation on 381 human day 3 embryo images.

Main Results:

• The developed method achieved a 0% error rate for 21.59% of embryos.
• 83.16% of embryos had a false recognition count of 2 or fewer blastomeres.
• The method demonstrated efficiency and speed in blastomere counting without requiring 3D reconstruction.

Conclusions:

• The proposed least square curve fitting (LSCF) approach offers a simple, efficient, and rapid solution for automatic blastomere counting from single embryo images.
• This method can aid in objective and reliable embryo viability assessment.