Artificial intelligence used to identify bacteria

Posted Jan 5, 2018 by Tim Sandle
Microbial identification has been streamlined in recent years through rapid methods and computer reading. However, a skilled microbiologist is often required. Can AI replace the need for the microbiologist?
This key experiment shows the successful protection of a phage-sensitive bacterial strain against a ...
This key experiment shows the successful protection of a phage-sensitive bacterial strain against a virus. Top-right - bacteria growing in the absence of a virus; Top-left - holes in the culture caused by an infecting virus; Bottom - when equipped with specific CRISPR defense system components, the bacteria became resistant to the virus.
John van der Oost
In many laboratories, from clinical to pharmaceutical, there is a shortage of microbiologists trained in identification - the process of determining one genus or species of bacterium or fungus from another. Perhaps, Beth Israel Deaconess Medical Center researchers contend, artificial intelligence can address this shortfall.
In the new research, the scientists have experimented with microscopes enhanced with artificial intelligence. These are designed to assist microbiologists diagnose microorganisms. The technology has been developed with the medical microbiology community in mind.
In trials, the showed how an automated artificial intelligence-enhanced microscope system was "highly adept" at identifying images of bacteria quickly and accurately. This was undertaken by training a convolutional neural network, a type of artificial intelligence modeled on the mammalian visual cortex, in order to categorize bacteria based on their shape and distribution. The images were of microscope slides where bacteria had been stained using the Gram-stain.
The Gram stain method employed includes the four-step technique: Crystal violet (primary stain); iodine (mordant); alcohol (decolorizer); safranin (counter stain) or the three-step method in which the decolorization and counter-staining step are combined. Done correctly, Gram-positive organisms retain the crystal violet stain and appear blue; Gram negative organisms lose the crystal violet stain and contain only the counter-stain safranin and thus appear red.
The morphological characteristics were selected to represent bacteria that most often cause common infections, including the rod-shaped bacteria including Escherichia coli; the cocoidal clusters of Staphylococcus species; and the pairs or chains of Streptococcus species. In all, some 25,000 images were used to train the platform. The success rate was put at 95 percent.
According to lead researcher Dr. James Kirby: ""This marks the first demonstration of machine learning in the diagnostic area. With further development, we believe this technology could form the basis of a future diagnostic platform that augments the capabilities of clinical laboratories, ultimately speeding the delivery of patient care." One advantage here is that images can be sent remotely and read by the artificial intelligence anywhere in the world.
The research has been published in the Journal of Clinical Microbiology. The research paper is called "Automated Interpretation of Blood Culture Gram Stains using a Deep Convolutional Neural Network."
Importantly, while artificial intelligence may be able to replace or supplement the role of the skilled microbiologist in deterministic microbiology a microbiologist is still required to ascertain if the obtained result is probable (all identification systems have inherent weaknesses) and to know what the implication of the obtained result is.