In publishing, a picture is worth a thousand words, as the saying goes. In bioprocessing, though, a picture could be worth even more. Even with all of today’s sensors, some steps still involve some form of imaging.
For example, tracking the concentration of cells in a bioreactor making a biopharmaceutical often requires collecting a sample for staining and cell counting. That process could be turned into an in-line and automated process, according to work by Laura Breen, PhD, a biotechnologist in the cell technology group at the National Institute for Bioprocessing Research and Training (NIBRT) in Dublin, Ireland, and her colleagues.
Traditionally, cell concentrations in a bioreactor are analyzed by manually collecting a sample, staining it with trypan blue die to distinguish dead cells from live ones, and then using a cell counter. Instead, Breen and her colleagues described a method based on flow-imaging microscopy using the PharmaFlow imaging system.
Automatic cell selection
These scientists developed a method that automatically collects a sample of cells, dilutes it, and images are collected at 30–60 frames per second as the sample flows past the camera. Plus, CantyVision imaging analysis software collects 39 morphological descriptors that distinguish live and dead cells. Breen’s team reported that their “analyses indicate that the PharmaFlow measurements correlated quantitatively with trypan blue for cell density and viability.”
Various reasons drive the pharmaceutical industry to track the health of cells in a bioreactor as easily and automatically as possible. For example, as Breen and her colleagues pointed out: “There are inherent problems associated with offline analysis caused by time delays that prevent immediate bioprocess intervention and remediation.”
With an in-line method based on imaging and automated analysis, a bioprocessor can track cell viability and density as needed. In fact, such a system can provide even more information. “Multivariate data analysis can be applied to large data sets to enable ‘deep learning’,” Breen’s team noted. “Such analysis of cell populations may provide a real-time understanding of the dynamics of sub-populations and the possibility of intervention when there are signs of decreasing cell health.”
