Blog

How Can AI Improve Veterinary Medicine?

veterinarian and dog look at X-ray
Sep 4, 2020 at 3:43 pm
Conor Guptill

Anyone who has worked in a busy veterinary practice knows the feeling of too much to do and not enough time to do it. Sometimes an extra set of hands, or eyes, can make all the difference… just to help stay on top of things.

What if computers could be trained to think like humans, creating an extension of the veterinary team to get more done in less time?

That’s the power of artificial intelligence (AI), which is showing promising results in a few areas of veterinary medicine, including radiology, lab work, and diagnostic research.

Let’s take radiology, for example. Veterinarians know that the industry has a big problem: too many radiographs and not enough specialists to review them in a timely manner. Practitioners who send rads out for specialty reads often wait several days—and even weeks—for results. And that’s not expected to improve any time soon. According to a recent study, radiograph caseloads are expected to triple by 2022, and if that happens, 66% of the caseload will not be met. (“Veterinary Telemedicine: A System Dynamics Case Study” John Voyer and Tristan Jordan published February 15th, 2018.)

Waiting days or even weeks for radiology results means conditions go undiagnosed and treatment is delayed, which hurts both patient health and practice revenue. Sometimes patients pass away during the waiting period, or clients lose the sense of urgency to begin treatment because too much time has passed.

That’s where artificial intelligence can help. AI is software based on algorithms built by veterinary radiologists and trained to interpret abnormalities in a patient’s radiograph. Today, this AI software can be tied in with a veterinary practice’s PACS so that all radiology images, once uploaded, can be automatically scanned and evaluated. Results are available within five minutes, allowing the veterinarian to evaluate the findings, provide diagnoses, and start treatment while the client is still in the office. The best part of AI is its ability to screen large numbers of cases in a short amount of time with very consistent results.

veterinarian looks at X-ray

It almost sounds too good to be true, and the industry has raised several questions about the use of AI in radiology. Here are the questions we hear most often.

Why aren’t there enough radiology specialists to keep up with the workload?

That’s due to a combination of standards of care and the lack of radiology departments in veterinary schools. The trend for many corporate groups and emergency and specialty hospitals is to have boarded radiologists doing over-reads for all radiographs taken. Plus, younger veterinarians are not taught to read radiographs in veterinary school these days—instead they are taught to send out for specialty reads. Unfortunately, there aren’t enough board-certified radiologists to keep up with the demand. Today there are less than 35 veterinary radiology programs in the U.S., each graduating between 30 and 40 specialists each year.

Does AI replace radiology specialists?

Definitely not. AI depends on radiology specialists and the entire veterinary industry to continue to “learn.” AI and specialists work best when they work together, and the quality and accuracy of AI is totally dependent on the knowledge and skills of the specialists who train it. The more AI is trained, the more accurate it becomes. The best use of AI is to do a quick initial screen and alert when a specialist read is required.

What are the error rates? Is AI ever wrong?

Because radiographic AI is trained by humans, the error rate is about equal to the human error rate. Over time, an algorithm can be trained to be as good as a human radiologist, which has been documented in human medicine. (Algorithms in human radiology have their sensitivities and specificities documented and are typically in the high 80s and low 90s.)

Is AI expensive or cost prohibitive?

It’s actually less expensive to employ AI than to send all radiographs out for specialty reads. Specialists charge between $55 and $100 per read, while AI software costs $60 per month and includes up to 500 reads. The most profitable way to add radiographic AI is as follows:

  • Incorporate AI reads into the workflow so that all radiographs get an AI scan.
  • Have the veterinarian view and validate the AI read, share it with the client during the same appointment, and begin treatment.
  • For complex or unusual cases, send the radiograph out for a specialty read when the veterinarian deems it necessary.

This can end up improving revenue for the veterinary practice because client dollars that would have otherwise paid for a specialty read can be used for patient treatment, and a shorter turnaround time leads to increased opportunities for treatment (and revenue) when the client is still in the practice and can more easily say “yes.”

How does AI help specialty radiologists?

This gets back to the “too much work and not enough time” problem. Veterinary staff aren’t the only ones who feel that crunch. Radiology specialists do, too, since there aren’t enough specialists to keep up with the caseload.

The specialty radiologists who we’ve worked with love the idea of using trained AI technology to augment a veterinarian’s own decision process in a way that saves time and decreases the specialist workload. AI can also help prioritize work for the radiologists. For example, the AI system can be configured to identify STAT cases based on the radiologist’s own criteria, so once radiologists start their day, AI can provide a list of cases in the order that they should be reviewed.

The following two tabs change content below.
Conor Guptill
Conor Guptill is senior product manager at Patterson Veterinary Supply, where he oversees the MarketHound suite of technology solutions for veterinarians, including the Vetology AI service.
Conor Guptill

Latest posts by Conor Guptill (see all)


Speak Your Mind

*