Towards Fluorescent-Tag-Less Viral Titration: Automated Estimation of Cell-Size Distribution and Infection Level from...

Area of Science:

• Biotechnology
• Cell Biology
• Machine Learning

Background:

• Manual segmentation and size analysis of insect cells for infection assessment are laborious and time-consuming.
• Fluorescent imaging requires protein tagging, limiting its application in recombinant protein production and vaccine development.

Purpose of the Study:

• To develop and validate a deep learning-based method for automated detection and size analysis of infected insect cells using phase contrast imaging.
• To assess the potential of this method for classifying infection levels without fluorescent markers.

Main Methods:

• Compared Histogram of Oriented Gradients (HOG) + Support Vector Machine (SVM), Faster Region-based Convolutional Neural Network (Faster RCNN), and YOLO for cell detection and size calculation.
• Utilized transfer learning on a limited dataset for model training.
• Validated the difference in cell size distribution between infected and uninfected cells.

Main Results:

• YOLO demonstrated superior performance in cell detection and size calculation compared to HOG+SVM and Faster RCNN, especially with limited data.
• A significant difference in cell size distribution was observed between infected and uninfected cells.
• The YOLO-based approach successfully enabled classification of infection levels.

Conclusions:

• Deep learning, specifically YOLO, offers an efficient and accurate method for automated insect cell analysis from phase contrast images.
• This technique facilitates infection level classification without the need for fluorescent tagging, streamlining processes in biotechnology and vaccine development.