Research on cell detection method for microfluidic single cell dispensing.

Area of Science:

• Biotechnology
• Computer Vision
• Cell Biology

Background:

• Traditional single-cell dispensing methods like limiting dilution and fluorescence-activated cell sorting (FACS) have limitations.
• FACS and microfluidic methods using fluorescence detection can impact cell viability.
• Accurate single-cell isolation is crucial for various biological applications.

Purpose of the Study:

• To develop a nearly non-destructive single-cell dispensing technique.
• To improve the accuracy and efficiency of single-cell detection.
• To overcome the limitations of existing cell isolation methods.

Main Methods:

• Implementation of a single-cell dispensing method based on an object detection algorithm.
• Development of an automated image acquisition system for cell detection.
• Utilization of the PP-YOLO neural network model with ResNet-18vd backbone for image analysis.

Main Results:

• The PP-YOLO model, with ResNet-18vd, achieved 98.6% precision in detecting single cells.
• The model demonstrated a fast inference time of at least 0.9 ms per 320x320 pixel image.
• A balance between detection speed and accuracy was successfully achieved.

Conclusions:

• The proposed object detection-based method offers a highly accurate and efficient approach for single-cell dispensing.
• This technique minimizes potential damage to cells during the detection process.
• The developed system provides a promising alternative for precise single-cell isolation in research.