Deep-Learning-Based Polar-Body Detection for Automatic Cell Manipulation.

Area of Science:

• Biomedical Engineering
• Computer Vision
• Cell Biology

Background:

• Polar body detection is critical for automated cell manipulation.
• Existing methods struggle with polar body defocus and deformation during 3D cell rotation.
• Varying polar body sizes complicate detection.

Purpose of the Study:

• To develop a robust deep learning framework for polar body detection during cell rotation.
• To address challenges of defocus, deformation, and size variation in polar body detection.
• To improve the accuracy and reliability of automated cell manipulation processes.

Main Methods:

• A deep learning framework based on image segmentation was proposed.
• An improved U-net convolutional neural network (CNN) architecture was utilized.
• A specialized image transformation method was designed to simulate cell rotation conditions, including deformation.

Main Results:

• The proposed method achieved a high detection accuracy of 98.7% on a dataset of 1000 images.
• The framework demonstrated effective detection of defocused and variably sized polar bodies.
• The method performed well in simulated cell rotation scenarios, accurately identifying deformed polar bodies.

Conclusions:

• The deep learning framework offers a significant advancement in polar body detection accuracy.
• The method effectively handles challenges posed by 3D cell rotation, including defocus and deformation.
• This approach has broad applicability for various automated cell manipulation tasks.