Veterinarians and researchers at the University of California, Davis have developed a new way to detect leptospirosis, a life-threatening bacterial disease, in dogs using artificial intelligence.
Leptospirosis is caused by the Leptospira bacteria, according to American Veterinary Medical Association, and it is typically found in soil and water. Infection in dogs can result in kidney failure, liver disease and bleeding in the lungs, with early detection being a matter of life or death, UC Davis said in a news release.
"Traditional testing for Leptospira lacks sensitivity early in the disease process," said Krystle Reagan, lead author of the study and a board-certified internal medicine specialist, in the release. Detection of the disease can take more than two weeks, she said, because the test needs to indicate the level of antibodies increasing in a blood sample.
"Our AI model eliminates those two roadblocks to a swift and accurate diagnosis."
By predicting this disease earlier, doctors and pet owners can better understand the course of the disease and outcome, Reagan said.
What is Leptospirosis?
Infections stem from urine-contaminated soil, food, bedding or from an animal bite. Dogs can be exposed to the bacteria from drinking water in rivers, lakes and streams, or being in contact with infected wildlife, farm animals, rodents and other dogs.
Symptoms can vary and some pets might not show any at all, according to the Centers for Disease Control and Prevention. Typical signs include fever, vomiting, diarrhea, severe muscle pain and weakness.
If untreated, Leptospirosis can be fatal. But with prompt detection and treatment, 90 percent of dogs can overcome the disease.
Leptospirosis is a zoonotic disease, meaning it can spread from animals to humans. But most people will get it from water-related activities, rather than from an infected pet, according to the veterinary association.
How does the UC Davis AI model work?
The model was created by looking at data from routine lab work from more than 400 canine patients who were tested for the disease at the university's Veterinary Medical Teaching Hospital.
Reagan said in a phone interview that the AI model used complex statistical methods to look for patterns associated with an outcome in the blood work. Through this, researchers created a system where they can apply new lab work and make a prediction about whether the patient is infected.
It was then used to test a group of new dogs, which it correctly identified the nine dogs that were positive for the disease and the 44 that were negative.
What this means for vets and pet owners
According to the release, the purpose of the model is for it be an online resource where veterinarians can input their patient's data and get a detection result on time.
Reagan said she hopes there will be a type of web application or a system that integrates into commercial labs that run blood work, and that will flag and alert vets when a patient appears to have an infection. The veterinarian can then check with the dog and owner, and test for it.
"There's a lot more information that might be hidden in the bloodwork that we're doing, or the lab work that we're doing, that may not immediately catch the eye of veterinarians," Reagan said. "So artificial intelligence, machine learning might help us get even more information out of the lab tests that we're already doing by finding these types of hidden patterns."
She added that this type of technology could expand beyond leptospirosis and can be applied to many different diseases, helping enhance doctors' clinical decisions and identify diseases earlier.
How to treat leptospirosis in dogs
Leptospirosis in dogs is treated with antibiotics, dialysis or hydration therapy. According to the CDC, early treatment will help them recover quickly and reduce the severity of organ damage.
The agency said that the disease typically develops 5 to 14 days after exposure to bacteria, but can be shorter or take longer -- up to a month or more.