When a disease is spreading, a system that examines data from animal rehabilitation centers could provide early warnings. A troubling trend emerges, with marine birds dying from domoic acid toxicity triggered by dangerous algal blooms along the California coast. An early indicator of when and where the problem will spread is: Rescued California brown pelicans, red-throated loons, and other species are showing signs of neurological disorder at wildlife rehabilitation clinics. Despite their presence on the state map, these facilities are not sufficiently networked to address the problem. When one center’s staff diagnoses a sick bird, others 40 miles away may not have access to that information. As a result, UC Davis researchers recently developed an early detection system that employs artificial intelligence to classify admissions to rehabilitation centers in the hopes of alerting wildlife authorities and researchers to developing concerns among marine birds and other creatures. Intake reports from 30 California centers are scanned by their system, including the animal’s species, age, the reason for admission, and diagnosis. The AI then categorizes the data using natural language processing, looking for trends in the frequency of admissions associated with specific illnesses and injuries.
The researchers analyzed more than 200,000 recordings and five years of data to develop baselines for how frequently certain circumstances occur in the real world. When the system finds an abnormality, such as an unusually high number of instances in a certain species, it alerts wildlife experts via the system dashboard, email, or text message. Because the system can process rehab center admission data in just a day or two, it can send out “prediagnostic” notifications, which are more efficient than waiting for diagnoses to be confirmed.
The researchers published an article in the journal Proceedings of the Royal Society in July reporting a test of their technology.“We wanted to use the data in an aggregate form to better help rehabbers to see the bigger picture, other than what they see at their individual centers,” President of the Wild Neighbors Database Project and one of the paper’s authors, Devin Dombrowski, says.
During the one-year pilot study, the system discovered a number of patterns that pointed to more serious issues. An inflow of marine birds with neurological signs such as twitching of the head and whole-body tremors prompted a warning. These birds, including the black and white waterbird species western grebes, were discovered to have domoic acid poisoning after postmortem examination. A high rate of clinic admissions for rock pigeons with neurological disorders in the San Francisco Bay Area had generated another notice a few months ago. The parasite Sarcocystis calchasi was discovered to be the source of the problem. Dombrowski and his wife, Rachel Avilla, are both wildlife rehabilitators who launched the Wild Neighbors Database Project in 2010 with the goal of standardizing record-keeping. They were frustrated that rehab center workers were tracking medical records in a variety of ways, from paper records to separate Excel spreadsheets.
The AI-based alert system at UC Davis is founded on the foundations of this older database. (The Wildlife Morbidity and Mortality Event Alert System is the name of the new system.) The system only took data from 30 California sites that are already WRMD contributors for the pilot study, although the WRMD database contains over 2 million historical records in total. It’s because of this that the researchers were able to access documents dating back five years. The study authors hope to unveil an enhanced version of the alarm system by the end of 2021, which will be tested by the California Department of Fish and Wildlife. For the time being, they’re refining the machine learning model through a process known as retraining, which involves feeding fresh data into the old pipeline in order to improve the model’s accuracy. The researchers are also working on novel algorithms that can predict several clinical classifications, such as cases when both neurological and ocular diseases coexist. Once everything is operating properly, the study authors hope to expand outside California, forming networks to aid wildlife agencies and veterinarians in other states, over the next few years.
“The methodology is flexible enough to accommodate different regions and different kinds of animal taxa,” says Pranav Pandit, a postdoc at UC Davis who developed the study’s mathematical models. Purdin, a veterinarian in Los Angeles County, expects that the alert system will assist veterinarians in preventing illness epidemics.
