Artificial Intelligence in Veterinary Medicine
Whether you regard artificial intelligence (AI) as an existential threat to humanity or the foundation of a Brave New World of limitless progress and prosperity, there’s no doubt it’s here to stay. And despite recent bipartisan calls to regulate its development and applications, there’s a good chance it won’t be restrained for very long.
What does that mean for veterinarians? Should you worry that virtual veterinarian chatbots in the service of giant internet entities could soon replace you? Probably not. While AI programs capable of using and understanding everyday natural language already exist, in the near term these would likely be used to dispense the kind of advice already available to pet parents in print form on countless websites (Does my dog have arthritis? Why is my cat’s breath terrible?). For veterinarians the technology may in fact be an opportunity to practice ultimate best medicine with the help of an electronic consultant of unfathomable knowledge and experience.
But that won’t be the end of it. AI is already making inroads in radiology. Currently there are not enough radiologists to read the countless radiographs taken daily in general practice, which can result in delayed diagnoses and treatment. While AI hasn’t advanced far enough to take over the radiologist’s role (at least not yet) it’s already being used make the process of interpreting scans and films more efficient. AI programs trained on many thousands of radiographs and the interpretations associated with each can use these data to make diagnoses of their own when shown radiographs to which they had not previously been exposed. Exactly how AI comes to its diagnostic conclusions isn’t always obvious (indeed, the machine’s thought processes can be downright mysterious) but its findings are sufficiently accurate to serve as a first pass review to help speed the radiologist’s work.
AI’s technology has only scratched the surface on predicting diseases. AI algorithms can analyze data from electronic health records, including medical history, breed, age, and environmental factors, to predict the likelihood of an animal developing a particular disease. Many diseases have vague clinical signs and may go undetected for years. This includes Addison’s disease also called the “Great Pretender” because its signs that mimic those kidney and intestinal disease. Drs. Krystle Reagan, DVM, PhD, DACVIM (SAIM) and Chen Gilor, DVM, PhD, DACVIM (SAIM) teamed up with an electrical and computer engineer at UC Davis to develop an AI algorithm to diagnose Addison’s disease with an accuracy rate greater than 99 percent. The algorithm was based on bloodwork for more than 1,000 dogs previously treated at UC Davis.
For oncologists and internal medicine veterinarians, diagnosis is not the ultimate goal. Personalized medicine and treatment options are the focus. AI is currently being used to develop custom treatment plans by analyzing an animal’s medical history, genetics, and other factors to recommend the most effective care. For example, Dr. Reagan’s research has shown that AI can help determine whether dialysis should be used as part of the treatment protocol for dogs diagnosed with leptospirosis.
How will veterinary medicine be influenced by AI in the future? ChatGBT can mimic speech surprisingly well today. AI in the future may be a sort of backup system to veterinarians in which telemedicine visits now can be triaged by an AI bot that determines a diagnosis and treatment plan and signed off by a human veterinarian thousands of miles away. You think we have Veterinary Client Patient Relationship (VCPR) issues now with telemedicine?
Artificial Intelligence, however, is not without its faults. When someone tells an untruth, we call it lying. When AI does it, it’s a hallucination. Part of the issue here is that we don’t completely understand how AI does what it does. Its database is the public domain so how (or why) can it make up fake Kings or history that never occurred? When while we are waiting for that answer, there are ways to limit the hallucinations. One of the easiest ways is limiting possible outcomes by specifying the response you want i.e., giving the AI bumpers in bowling.
AI has vast implications in veterinary medicine. AI can facilitate more comprehensive results in a shorter amount of time. Leveraging technology can also help the health of your practice too. Allowing technology to do what it is capable of allows you to diagnose disease earlier, make better treatment decisions, improve pet owner communication, and make that pet parent a client of your practice for their lifetime. Better science leads to better decisions, and better business. At least, it always has.
