Algorithms trained to interpret the data from microscopes can dramatically expand the information that can be derived from the resulting images, and even optimize the way imaging experiments are conducted. In this article, Nature Methods offers an overview of approaches developed in the field, by means of a series of discussions with experts in the field.

Our researcher Flavie Lavoie-Cardinal, member of the CERVO Research Center, captured the author’s attention when she described one of these platforms in a publication in 2019. She had described the use of machine learning to optimize STED imaging experiments involving a range of parameters, such as laser power and scan speed. After teaching the software what a “good” image looks like for a particular set of experiments, she found that she could quickly allow the microscope to set up on its own.

“The algorithm has gotten much better at finding the optimal settings for a workflow than an inexperienced user,” Lavoie-Cardinal said. “I was faster because I have ten years of experience behind me, but the algorithm was able to find the same parameters as those that I found.”