Category: Automation and High-Throughput Technologies

1148-D - Accuracy and Automatic Error Detection in Machine Vision Monitoring of Bulk Reagent Dispensing

Tuesday, February 6, 2018
5:00 PM - 6:00 PM

Automated liquid dispensing is a key step in a wide variety of biological and chemical assays.  We monitor bulk (microliter-scale) automated reagent dispensing using a high-speed line camera and process the recorded data into per-well volumes.  The technique is essentially independent of liquid type and is readily adaptable to all standard well plates.  A per-well volumetric error of 5% (rms) has been achieved across a range of dispensed volumes (5 to 300 µL), dispensing rates, and multiwell plate types (96 to 1536 wells).  The error is evaluated by comparing to reference measurements using a fluorescence plate reader, with additional corroboration based on gravimetric data for higher volumes.  The effect of liquid viscosity on volumetric accuracy is demonstrated to be negligible below a threshold value.  We also present example time-resolved imaging of dispensing failures such as exhaustion of the reagent supply and clogging of dispensing tips.  Running out of dispensed reagent is typified by the formation and discharge of small bubbles and smaller droplets, which sometimes occlude the view of the dispensing region, but can themselves be identified and used to determine that a dispensing error has occurred.  A survey of these types of errors demonstrates an extremely high degree of identification accuracy.  The clogging and spontaneous un-clogging of dispensing tips, important during dispensing of certain proteins or cells, are also confirmed to be easily identifiable.

Alan Hoskinson

Principal Scientist
Physical Sciences Inc.
Andover, MA

Applied physicist with expertise in spectroscopy, image analysis, plasma science, and microwave circuits.