Battling the radiology backlog

The use of AI in veterinary radiology can make both radiology specialists and veterinarians more efficient

Photos courtesy Vetology
Photos courtesy Vetology

During a particularly hectic day, when there’s not enough time to get everything done, don’t you sometimes wish you could clone yourself? Or when veterinary specialists are backlogged with cases, do you wish you could clone them?

Veterinary radiology is one specialty that is always backlogged. According to the American Veterinary Medical Association (AVMA), the demand for advanced remote radiology is exceeding the supply of board-certified radiologists, and that trend is expected to continue over the next several years. Experts predict a 300 percent increase in teleradiology caseloads, which will overwhelm the current number of specialists. A 2018 study predicted 66 percent of the North American veterinary teleradiology caseload will not be met by the end of 2022.1

Cloning may not be available today, but artificial intelligence (AI) is. Using it effectively in veterinary radiology can make both radiology specialists and veterinarians more efficient, improving patient care and practice revenue.

How did we get here?

How did the industry get to a place where the radiology caseload greatly outweighs the number of specialists available? The answer is lack of radiology departments and current standards of care.

The number of veterinary schools without radiology departments is increasing, which translates to more graduating veterinarians with less training on how to interpret radiographs. Schools without radiology programs, and even some with, send their radiographs to specialists for review. Graduating veterinarians learn to do the same. This compounds the teleradiology caseload.

That wouldn’t be a problem if board-certified radiologists were plentiful, but unfortunately, they aren’t. Today, there are less than 35 veterinary radiology programs in the U.S., each graduating between 30 to 40 specialists every year.

Differing standards of care also play a role in the imbalance between cases and specialists. Some veterinarians feel comfortable reading radiographs and sending only complex cases to specialists. Others believe DVMs should not interpret radiographs, instead focusing on core competencies and leaving reading radiograph to specialists. Yet, the quantity of radiographs taken and sent out on a daily/weekly/monthly basis has grown in the past few years, thanks to less expensive technology and ease of radiograph transmission. As a result, there aren’t enough specialists to handle the load.

Ramifications for the industry

The trend for many large corporate groups and emergency and specialty hospitals is to have boarded radiologists doing overreads for all radiographs taken to follow human medicine procedures more closely. Unfortunately, as we continue to produce more radiographs than specialists can read, turnaround times continue to get longer and longer. Diagnoses, treatments, or further testing will also be delayed, ultimately compromising our patients’ health. Interestingly, fast turnaround time has become the new currency in radiology services, as it’s not unusual for a non-STAT turnaround time to reach 15 days or more.

Sending radiographs to specialists also increases the cost for pet owners. While that expense is justified, especially when a specialist read is critical to understanding a unique or complex case, this sometimes means the more money clients spend on specialist radiography reads, the less they have for follow-up treatment or future visits. This compromises not only the pet’s health, but also the veterinarian’s revenue and profitability.

How can technology help?

Increasing the number of radiology programs in the U.S. to train and certify more specialists is one solution. However, that takes time to implement and won’t help the industry right now, or even in the short term.
A more immediate solution would be to employ AI to automate routine radiology readings in the workflow, saving the more difficult cases for the limited supply of specialists.

Radiographs can be scanned quickly by artificial intelligence technology to flag any abnormalities.
Radiographs can be scanned quickly by artificial intelligence technology to flag any abnormalities.

Today, AI technology can be tied in with your software so that all radiology images, once uploaded, can be automatically scanned and evaluated. Results are available in minutes (not hours or days), allowing the veterinarian to evaluate the findings quickly, provide diagnoses and recommendations to the pet owner while he/she is still in the office, and begin treatment immediately.

Some veterinarians worry AI may replace radiology specialists, while others may not feel comfortable trusting radiograph interpretation to a “computer,” rather than a trained specialist. The truth is, AI and radiology specialists go hand in hand and work best together.

Radiographic AI is based on algorithms developed by veterinary radiologists and trained to reliably interpret any abnormalities in a pet’s radiographs. Think of AI as transferring radiologists’ knowledge and experience to a computer chip that doesn’t get tired by the end of the day. The quality of AI is fully dependent on the knowledge, experience, and skills of the specialists who train it initially and who continue to improve it over time.

How accurate are AI reads? When assessing a screening exam, high sensitivity and specificity are critical to be sure false negatives and false positives are maintained as low as possible. An AI algorithm should also have systems in place to test these metrics periodically to be certain accuracy is preserved. Different AI vendors, machine settings, and animal positioning are among the many things that can affect the output of an algorithm, so careful surveillance is important to maintain a safe product.

AI’s greatest strength is screening large numbers of cases in a short amount of time, with very consistent results, freeing up both veterinarians and radiology specialists to focus only on the screenings requiring human input.

No veterinarian has time to do everything, and no human catches everything 100 percent of the time, particularly on hectic work days. That’s when things slip through the cracks and radiographic findings can be missed. With ongoing work, an algorithm can be trained to be as good as a human radiologist, which has been documented in human medicine. Additionally, algorithms in human radiology approved by the U.S. Food and Drug Administration (FDA) have their sensitivities and specificities documented online and are typically in the high 80s and low 90s.

Better health, more revenue

Leveraging technology to speed up radiography results is a benefit for both patient health and practice revenue.

AI can be tied in with software so that all radiology images, once uploaded, can be automatically evaluated.
AI can be tied in with software so that all radiology images, once uploaded, can be automatically evaluated.

The patient health improvement aspect is obvious. Quicker, more reliable results mean faster diagnoses, better treatment, and healthier pets. Having an AI screening tool helps empower the practice overall, particularly in clinics with younger veterinarians who may not read radiographs fluently. Veterinarians can pay from $55 to more than $100 for a specialty radiologist to interpret radiographs, depending on the number of images and regions to read. On the other hand, some providers of radiographic AI charge a monthly fee that allows a certain number of radiograph rates per month.

In some cases, however, faster results can save lives. One veterinary practice thought a dog was in heart failure and the owners were contemplating euthanasia. A quick AI read of the radiograph said there was no evidence of heart failure. This result prompted the veterinarian to send the case in for review. AI was correct—the dog’s problem was caused by another, more treatable issue.

The revenue improvement is more surprising—shorter teleradiology turnaround times with AI can lead to increased opportunities for treatment and the corresponding revenue. For example, one veterinary practice in Los Angeles, Calif., used to wait three to five days for a radiograph to be read by a specialist. During that window, pets sometimes passed away, or clients lost the sense of urgency for treatment and didn’t return to the practice for the follow-up recommendations. Now, when AI radiology reports are automatically available in five minutes and recommendations are made while clients are still at the practice, clients are in a better place to agree to the next steps.

Using technology to lower our costs and leverage our time is simply good practice and good business. Most pet owners have a fixed amount of money to spend on pet care. They can pay the extra money for a specialty radiology read, or they can spend the money on treatment recommendations and regular visits. It makes sense to use technology to screen radiographs, diagnose and recommend treatment as soon as possible, and save the specialists’ time for the radiographs that truly need more professional evaluation.

At least, of course, until we can clone ourselves.

Seth Wallack, DVM, DACVR, is CEO and founder of Vetology, a cloud-based veterinary teleconsulting platform, and Vetology AI, which delivered the first commercial small animal radiology artificial intelligence system. A veterinary radiology specialist himself, Dr. Wallack has served as the director of the American Association of Veterinary Radiologists (AAVR) and CEO of the Veterinary Imaging Center of San Diego. He is also founder of DVMInsight, a veterinary teleconsulting company that began in 2005.

References

1 “Veterinary Telemedicine: A System Dynamics Case Study” John Voyer and Tristan Jordan published February 15, 2018.

Comments
Post a Comment

Comments

bocoran admin jarwo