Deep learning-assisted 10-μL single droplet-based viscometry for human aqueous humor

Abstract

Probing the viscosity of human aqueous humor is crucial for optimizing micro-tube shunts in glaucoma treatment. However, conventional viscometers are not suitable for aqueous humor due to the limited sample volume—only tens of microliters—that can be safely extracted without causing permanent ocular damage. Here, we present an artificial intelligence-assisted microfluidic viscometry for measuring 10-μL aqueous humor collected at the point of care. Our approach involves injecting a single droplet of the sample into a microfluidic chip using hydrostatic pressure, minimizing interfacial effects with surfactants and hydrophobic coatings, and analyzing the sample flow using a deep learning-based detection scheme. For the first time, we have measured the viscosity of a 10-μL human aqueous humor and observed approximately 30 % variation between individuals. These individual differences in aqueous humor viscosity should be considered when designing microtube shunts for glaucoma treatment. Our method paves the way for the viscometry of small-volume biofluids, enabling new diagnostic and therapeutic applications in biomedical technology.