Microscopes with artificial intelligence (AI) built in, could help clinical microbiologists diagnose potentially deadly blood infections. This could also improve patients’ odds of surviving these diseases, according to a microbiologist at Beth Israel Deaconess Medical Center (BIDMC).
The scientists demonstrated this process in an automated AI-enhanced microscope system that is, according to them, “highly adept” at identifying images of bacteria quickly and accurately. This automated system could help alleviate the current lack of highly trained microbiologists. The shortage of microbiologists is expected to worsen as 20 percent of technologists reach retirement age within the next five years.
"This marks the first demonstration of machine learning in the diagnostic area," said the senior author of the paper, James Kirby, MD, Director of the Clinical Microbiology Laboratory at BIDMC and Associate Professor of Pathology at Harvard Medical School. "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."
Kirby’s team used an automated microscope designed to collect high-resolution image data from microscopic slides. The team used blood samples taken from patients who were suspected to have bloodstream infections. The blood samples were incubated to increase bacterial numbers. The slides were then prepared by placing a drop of the blood on a glass slide and stained with a dye that makes the bacterial cell structure more visible.
Next, they trained a convolutional neural network (CNN). CNN is a class of artificial intelligence that is modeled on the mammalian visual cortex and used to analyze visual data. Using CNN, they categorized bacteria based on their shape and distribution. The characteristics were selected to represent bacteria that most often cause bloodstream infections: the rod-shaped bacteria, including E. coli; the round clusters of Staphylococcus species and pairs; or chains of Streptococcus species.
"Like a child, the system needed training," said Kirby. "Learning to recognize bacteria required a lot of practice, making mistakes and learning from those errors."
In order to train the system, the researchers fed the unschooled neural network more than 25,000 images from blood samples that were prepared during routine clinical workups. By cropping these images, which have already been identified by a human clinical microbiologist, the researchers generated more than 100,000 training images. The AI learned how to sort the images into three categories: rod-shaped, round clusters and round chains or pairs. Ultimately, the AI microscope achieved nearly 95 percent accuracy.
The team challenged the algorithm to sort the new images from 189 slides without human intervention. In the end, the algorithm achieved more than 93 percent accuracy in all three categories. With further development and training, Kirby and colleagues believe the AI-enhanced platform could be used as a fully automated classification system in the future.
Until then, Kirby believes that automated classification can ameliorate the shortage of human technologists by helping them work more efficiently.
This could "conceivably reduce technologist read time from minutes to seconds," said Kirby.
Human technologists routinely provide highly accurate diagnoses, and the demand for these workers is currently exceeding supply in the United States. Nine percent of lab technologist positions remain unfilled, and that number is expected to dramatically increase when the Baby Boomer generation of technologists begin to retire.
These images can be sent remotely, which brings the highest level of expertise anywhere the internet can reach. This is critical because rapid identification and delivery of antibiotic medications is the key to treating bloodstream infections, which can kill up to 40 percent of patients who develop them. Every day that a patient goes untreated or undiagnosed is linked to increased risk of mortality.
Along with the critical uses, the new tool can be used in microbiology training and research.
"The tool becomes a living data repository as we use it," Kirby said. "And could be used to train new staff and ensure competency. It can provide an unprecedented level of detail as a research tool."
The paper on this research was published in the Journal of Clinical Microbiology.