How Artificial Intelligence is Improving Radiology Workflows
Modern computing technology has improved efficiency in radiology in the same ways that it has in many fields. But many experts expected the benefits of IT on radiology to plateau. But now, the advent of real Artificial Intelligence is breaking down anticipated roadblocks, offering major improvements in the efficiency, quantity, and quality of the work radiologists do.
This technology is a part of what has been called the 4th industrial revolution. It’s an irreversible trend that has many worrying about lost jobs as machines do increasingly complex tasks. However, in radiology- a field where the number of new professionals is far below the demand for their skills, AI is making it possible for one radiologist to do more without making their skills obsolete.
Imaging Informatics professional from Emory University in Atlanta, Dr. Safdar, told RSNA.org, “The number of people who can interpret images can’t keep up with the growing demand for their skills- especially with an aging population. AI is helping providers to maximize their performance, make specific recommendations and help patients to better understand their particular situation.”
Trent Kyono, a lead author of a study published in the Journal of the American College of Radiology, writes, “[…] radiologists are under increasing pressure to deliver timely service. Because the large majority of mammograms a radiologist examines are negative, machine learning methods that triage a subset of examinations as negative with high accuracy … [are] freeing up time to focus on more suspicious examinations and diagnostic workups”
New developments in AI mean radiologists can spend less time verifying cases of non-disease conditions and spend more time spotting dangerous conditions.
AI is Improving Radiology Workflows in 3 Key Ways
According to Peter Eggleston, a global marketing director for GE Healthcare says, “AI is making radiology professionals more productive, boosting quantity, and enhancing the precision of their work.”
- Productivity: By automating and prioritizing routine tasks, AI is streamlining radiology workflows. This means less wasted time moving between discordant tasks as AI generates a more “connect the dots” task queue.
- Quantity: AI tools and applications can extract and quantify information either automatically or semi-automatically.
- Precision: Ensuring that the correct information is accessible, separated from non-useful information, and by ensuring that quantification processes are repeatable, AI improves accuracy.
AI Enhancement, Without Losing the Human Element
While these advanced tools offer significant advantages to radiology professionals, they are not reducing the value of the individual radiologist. Dr. Safdar, says, “These applications are not necessarily doing anything different from what radiologists already do. They just do it faster and without fatigue, making them useful in finding the proverbial needle in the haystack — the chest X-ray that shows a pneumothorax or the head CT that reveals a subarachnoid hemorrhage — and bringing it to the radiologist’s attention.”
For these types of systems to be effective requires technicians who can keep the negative predictive value (NVP) at a minimum of 99%. In one experimental system called AURA, 99% NVP was able to be maintained, decreasing the number of radiologists needed to diagnose positive cases by 34% in a simulated scenario where cancers occurred at a rate of 15%. This means radiology and IT can be expected to merge in the near future.
The authors of the AURA project wrote, “Our findings suggest the AURA system achieves this reduction by accurately classifying patients with attributes that are known to be related to lower probabilities of cancer, such as younger age and lower breast density.”
Such practical examples of the ways AI in radiology can free up more time for professionals to focus on cases with the greatest need is exciting. Radiologists will have the pleasure of providing better products to more patients without having their jobs threatened by automation. The only question left is exactly how much patient outcomes will improve.