Identifying parasites under the microscope has long been a painstaking task requiring highly trained experts to manually scour each sample for telltale cysts, eggs or larva. Now, a deep-learning model, known as a convolutional neural network (CNN), achieves that work with a high degree of precision, according to a study published Tuesday in the Journal of Clinical Microbiology.
The researchers demonstrated that the AI system can detect parasites in wet mounts of stool with greater sensitivity than human observers, even those with years of experience hunting for these signs.
"It has been a groundbreaking effort, and what we’ve accomplished is remarkable," said lead author Blaine Mathison, ARUP’s technical director of parasitology and an adjunct lecturer in the University of Utah’s Department of Pathology. "Our validation studies have demonstrated the AI algorithm has better clinical sensitivity, improving the likelihood that a pathogenic parasite may be detected."
A leading national reference lab, ARUP is an independent nonprofit enterprise of the University of Utah and the School of Medicine’s Department of Pathology, where Mathison is an adjunct instructor.
To build and test the system, ARUP and its partner, a Utah tech firm called Techcyte, trained the AI using more than 4,000 parasite-positive samples collected from laboratories across the United States, Europe, Africa and Asia. These samples represented 27 classes of parasites, including rare species, such as Schistosoma japonicum and Paracapillaria philippinensis from the Philippines, and Schistosoma mansoni from Africa.
"This was really a robust study when you consider the number of organisms and positive specimens used to validate the AI algorithm," Mathison said.
After discrepancy analysis, the positive agreement between AI and manual review was 98.6%. The tool also picked up 169 additional organisms that had been missed in earlier manual reviews.
"We are identifying more organisms than we would without the AI, which improves diagnosis and treatment for patients who are affected," said Adam Barker, ARUP’s chief operations officer.
Furthermore, a limit of detection study found AI consistently found more parasites than the technologists did, even when the samples were highly diluted, suggesting the system can detect infections at earlier stages or when parasite levels are low.
ARUP has pioneered the use of AI in clinical parasitology for years. In 2019, it became the world’s first lab to apply AI to the trichrome portion of the ova and parasite test. In March 2025, it expanded that capability to include the wet-mount analysis - becoming the first laboratory to use AI for the entire testing process.
That timing proved propitious: in August, ARUP received a record number of specimens for parasite testing. The efficiency gained through AI enabled the lab to meet demand without compromising quality.
"An AI algorithm is only as good as the personnel inputting the data," Barker said. "We have phenomenal staff who have used their extensive knowledge and skills to build an exceptional AI solution that benefits not just the laboratory, but also patients."
ARUP and Techcyte plan to continue expanding AI’s role in diagnostic testing. Beyond parasitology, ARUP has already implemented AI to assist with Pap testing and is developing other tools to streamline lab operations and improve diagnostic accuracy.
Mathison BA, Knight K, Potts J, Black B, Walker JF, Markow F, Wood A, Bess D, Dixon K, Cahoon B, Hymas W, Couturier MR.
Detection of protozoan and helminth parasites in concentrated wet mounts of stool using a deep convolutional neural network.
J Clin Microbiol. 2025 Oct 21:e0106225. doi: 10.1128/jcm.01062-25