Stain-Free Approach to Determine and Monitor Cell Heath Using Supervised and Unsupervised Image-Based Deep Learning.

Area of Science:

• Biotechnology
• Cell Biology
• Artificial Intelligence in Medicine

Background:

• Cell-based medicinal products (CBMPs) offer novel therapeutic potential for complex diseases.
• Current cell quality control methods are often labor-intensive, costly, and rely on specific staining assays.
• There is a critical need for rapid, robust, and non-invasive analytical techniques for CBMP manufacturing.

Purpose of the Study:

• To develop and validate a stain-free, deep learning-based method for cell health assessment in CBMPs.
• To utilize flow imaging microscopy (FIM) and cellular morphology for predicting cell viability and state.
• To explore the utility of unsupervised learning (variational autoencoders) for process monitoring in CBMP production.

Main Methods:

• Combined image-based deep learning with flow imaging microscopy (FIM) on unstained Jurkat cells.
• Developed a supervised fingerprinting algorithm trained on labeled images of healthy, dead, and apoptotic cells.
• Employed an unsupervised variational autoencoder (VAE) for morphological feature learning and sample classification.

Main Results:

• The stain-free method accurately predicted cell viability, showing good agreement with traditional stain-based assays.
• The supervised algorithm successfully distinguished between healthy, dead, and apoptotic cells, independent of specific treatments.
• The unsupervised VAE effectively classified cell states and identified cellular debris based on learned morphological features without human labels.

Conclusions:

• Deep learning combined with FIM provides a robust, non-invasive, and non-destructive method for cell quality control in CBMP manufacturing.
• Stain-free morphological fingerprinting offers a faster and potentially more cost-effective alternative to traditional methods.
• Unsupervised VAEs demonstrate significant potential as exploratory tools for real-time process monitoring in cell therapy production.