Rapid and sensitive mycoplasma detection system using image-based deep learning.

Area of Science:

• Biotechnology
• Cell Therapy Manufacturing
• Microbiology

Background:

• Cell therapy manufacturing costs are high, largely due to stringent quality control measures.
• Mycoplasma testing is crucial but time-consuming, expensive, and requires expert interpretation.
• Manual detection of mycoplasma contamination from images is labor-intensive and prone to error.

Purpose of the Study:

• To develop and evaluate a convolutional neural network (CNN) program for rapid and sensitive mycoplasma detection in cell cultures.
• To reduce the time and cost associated with traditional mycoplasma testing methods.
• To streamline the quality control process in cell therapy manufacturing.

Main Methods:

• Development of a three-part CNN program for mycoplasma detection, prediction, and cell counting.
• Training the CNN using stained DNA images of mycoplasma-infected and non-infected Vero cells.
• Comparison of the CNN program's performance against manual counting methods.

Main Results:

• The CNN program achieved a minimum detectable contamination level of 5 CFU, compared to 10 CFU for manual counting.
• The proposed program reduced testing time by 20-fold compared to manual observation.
• The system demonstrated improved sensitivity and efficiency in identifying mycoplasma contamination.

Conclusions:

• The developed CNN program offers a faster, more sensitive, and cost-effective solution for mycoplasma testing.
• This AI-driven approach can significantly contribute to a streamlined manufacturing process for cellular products.
• The system holds promise for improving quality control in both cell-based research and clinical applications.