Veterinarians at the University of California, Davis School of Veterinary Medicine have developed an algorithm utilizing artificial intelligence (AI) to detect Addison’s disease, a rare, life-threatening illness in dogs. Addison’s disease, also known as hypoadrenocorticism, is a condition that results in a lack of critical hormones, which are needed to maintain health.
UC Davis’ Krystle Reagan, DVM, PhD, DACVIM (SAIM) and Chen Gilor, DVM, PhD, DACVIM (SAIM) (with UC Davis at time of development; currently at the University of Florida) teamed with an electrical and computer engineer to develop the AI algorithm, which has an accuracy rate greater than 99 percent. The team touts the program to be superior to any other screening tool that utilizes routine blood tests available to veterinarians.
Addison’s disease is notoriously difficult to recognize. Dogs have vague clinical signs that mimic other conditions such as kidney and intestinal disease, causing veterinarians to refer to the disease as “The Great Pretender.” Addison’s can go undetected for years.
“Anecdotally, we see dogs with Addison’s disease come through the clinic, and they’ve been misdiagnosed for two to three years,” said Dr. Reagan, a board-certified small animal internist with the UC Davis veterinary hospital. “Once Addison’s is properly detected, though, it is generally easy to treat with an excellent prognosis for the patient.”
“We set out to create an alert system that uses information from routine screening tests,” Dr. Reagan continued. “The alert should be able to inform veterinarians when Addison’s disease is likely, and prompt further investigation.”
Their AI-powered algorithm does just that. When a sick dog visits a veterinarian, often the first tests ordered are routine blood tests (complete blood count and serum biochemical profile). The loss of hormones associated with Addison’s disease results in subtle irregularities in those tests that can be confused with other disease processes. The team used this routine blood work to train an AI program to detect complex patterns from more than 1,000 dogs previously treated at UC Davis. The computer program was able to learn these patterns, and with very high accuracy, determine if a dog has Addison’s disease.
This program can analyze this first line, routine blood work, and alert veterinarians when Addison’s disease is suspected, triggering them to pursue further diagnostic testing – an adrenocorticotropic hormone stimulation test (the gold standard to confirm Addison’s).
The team has filed a non-provisional patent through the UC Davis Office of Research and has a commercialization plan in place to license the program to large laboratories whose services are used by most veterinary practices. The program is anticipated to be available for commercial use by the end of 2020.
“Veterinarians need a safety net to prevent dogs with Addison’s from falling through the cracks,” said Dr. Reagan. “This AI program is now that safety net. It has the potential to revolutionize the detection of Addison’s and save many dogs’ lives.”
As Addison’s disease also affects many humans, the team is hoping for a translational component of this to help people with this disease. Dr. Reagan is currently collaborating with physicians and researchers to increase the utilization of AI to advance human medicine.
The UC Davis Center for Data Science and Artificial Intelligence Research (CeDAR) recently hosted Health Sciences Data Day, which Dr. Reagan attended and shared data and visionary ideas with physicians and human medicine researchers on subjects such as AI, machine learning, and other benefits of health sciences data collection and analysis.
CeDAR is committed to unleashing the true effectiveness of data science and AI through interdisciplinary collaboration. As a part of the UC Davis health sciences system (that includes the medical, veterinary, and nursing schools), CeDAR has the opportunity and resources to bring together world renowned experts from many fields of study with top data science and artificial intelligence researchers. As CeDAR advances data science foundations, methods and applications, it weaves them into the fabric of the university, promoting a highly efficient exchange of information and expertise that enhances real-world data science applications.
This AI breakthrough has much potential to optimize detection of other veterinary diseases as well. Dr. Reagan’s research is showing great promise for the early diagnosis of leptospirosis in dogs – detecting the disease earlier than any other tool to help give pet owners a prognosis when making difficult decisions surrounding dialysis. The current tests for leptospirosis do not work well early in the infection when veterinarians and pet owners need to make critical decisions about dialysis.
“There is a lot of other exciting research currently surrounding AI looking at prediction of adverse events, identifying diseases, and monitoring disease treatment,” Dr. Reagan said.
Dr. Reagan and her colleagues recently published this Addison’s breakthrough in the Domestic Animal Endocrinology journal, and she delivered a presentation on the topic at the 2019 American College of Veterinary Internal Medicine’s annual conference. She also presented the research at the 2019 UC Davis Veterinary Intern and Resident Research Seminar, where it was awarded the grand prize.
# # #