Deep learning software helps identify tiny bacteria in microscopy images

Omnipose, a deep learning software, helps solve the challenge of identifying varied and tiny bacteria in microscopy images. He went beyond this initial goal to identify several other types of tiny objects in the micrographs.

The UW Medicine Microbiology Lab of Joseph Mougous and the University of Washington Physics and Bioengineering Lab of Paul A. Wiggins tested the tool. It was developed by University of Washington physics graduate student Kevin J. Cutler and his team.

Mougous said that Cutler, as a student of physics, “demonstrated an unusual interest in immersing himself in a biological environment so that he could learn first-hand the problems to be solved in this field. He came to my lab and quickly found one that he solved spectacularly.”

Their results are published in the October 17 edition of Natural methods.

Scientists found that Omnipose, trained on a large database of bacterial images, worked well in characterizing and quantifying the myriad of bacteria in mixed microbial cultures and eliminated some of the errors that can occur in its predecessor, Cellpose. .

Moreover, the software was not easily fooled by extreme changes in a cell’s shape due to antibiotic treatment or antagonism by chemicals produced during interbacterial attack. In fact, the program has shown it can even detect cell poisoning in a test using E. coli.

In addition, Omnipose did well in overcoming recognition problems due to differences in optical characteristics between various bacteria.

Most bacteria are spheres or rods, but some have other basic shapes, such as twisted spirals. In addition to this, Omnipose could identify more elaborate bacteria with elongated shapes or with branches, filaments and appendages, all physical traits that can make it difficult for deep learning tools to determine which bacteria are present in a picture.

The program still faces some limitations in handling object overlap in a 2D interpretation of a 3D sample of a crowded microbial community. The overlapping of objects is what produces, for example, the effect of a clock on a wall giving the illusion of sticking out of a person’s head in a photograph.

By analyzing the cells of a root primordial data set of the fast-growing weed A. thalianaOmnipose nevertheless showed some advantages over previous approaches in this 3D sample.

Further examinations by the Mougous lab team of Omnipose’s capabilities showed that bacteria below a certain size threshold can be difficult for the tool to detect.

Despite these drawbacks, the researchers believe that Omnipose could be a solution, they noted, to “help answer various questions in bacterial cell biology.”

To see if it could also become a multifunctional tool in other biological or even non-living areas dependent on microscopy, the scientists tested the program on micrographs of the ultra-tiny roundworm. C.elegans, an organization important in research on genetics, neuroscience, development and microbial behavior. Like some bacteria, this creature has an elongated shape. Like many other worms, it can also contort. Omnipose could choose C.elegans regardless of its various stretches, contractions and other movements. This ability could be useful, for example, in neural studies of C.elegans locomotion during accelerated tracking.

By designing tools like Omnipose, researchers are investigating a pixel-precise scale to define the boundaries of a cell. Indeed, most images of bacterial cell bodies are composed of only a small number of pixels. The researchers explained that defining boundaries within an image is called segmentation. They developed Ominpose through a deep neural network high-precision segmentation algorithm. Their experiments showed that Omnipose has unprecedented segmentation accuracy.

The scientists designed Omnipose for use by typical research labs and made its source code, training data, and models publicly available, along with documentation on how to use the program.

“We anticipate that Omnipose’s high performance on diverse cellular morphologies and modalities,” the researchers wrote in their report, “could unlock insights from microscopy images that were previously inaccessible.”

Reflecting the importance of the problem, this is a cluttered area. Still, Kevin’s solution stands out. We believe this will be a game-changer for biological image analysis.”

Joseph Mougous, UW Medicine

In addition to Cutler, Wiggins and Mougous, the other Omnipose test project researchers were Carsen Stringer, Teresa W. Lo, Luca Rappez, Nicholas Stroustrup. S. Brooke Peterson and Paul Wiggins. Mougous is a researcher at the Howard Hughes Medical Institute.


Journal reference:

Cutler, KJ, et al. (2022) Omnipose: a high-precision morphology-independent solution for bacterial cell segmentation. Natural methods.

Comments are closed.