A new study published in Scientific Reports details the development of a machine learning (ML) model to accurately identify isolated cells, which may streamline future research processes and quicken the development of various treatments.
Cell isolation is a necessary step in cell biology research as clustered cells interfere with each other, skewing the results when trying to study a single type of cell. Identifying isolated cells can be challenging, however, and has always relied on humans to manually confirm if a cell has been isolated by examining the microscopic image. This reliance on human intervention has become a bottleneck in research processes.
Advanced Lab Management Certificate
The Advanced Lab Management certificate is more than training—it’s a professional advantage.
Gain critical skills and IACET-approved CEUs that make a measurable difference.
In an attempt to address this bottleneck, a team of researchers from Toyohashi University of Technology’s Department of Mechanical Engineering, as well as others from the Electronic Inspired Interdisciplinary Research Institute, developed a ML model that can reliably detect single cells in microscopic images, easing the burden on humans.
The team accomplished this by first trapping single cells in 30μm diameter microwells formed in a hydrogel, whose properties extended the observation period of cell behavior due to its biocompatible nature. They then sorted each cell by which had single cells and which didn’t. The team used the sorted images as the base dataset upon which to train a ML model to identify isolated cells.
In testing the model, the team found that it could correctly discern the presence of single cells with a mean Average Precision (mAP) of 0.801, out of a maximum possible score of one, in just 0.09 seconds when using bright-field microscopy. Looking to improve those metrics, the researchers tested the model with fluorescence-stained cell images, which had higher contrast than the bright-field images and allowed cells to stand out more clearly. The model’s performance improved, with its mAP jumping to 0.989 and its inference time dropping to 0.06 seconds.
This new technology for identifying isolated cells can shorten research turnaround time, quickening advancements in a variety of medical engineering applications like cancer diagnosis and drug discovery screening, which will contribute to the development of new treatments.
