The radiology artificial intelligence marketplace is growing quickly, with the Radiological Society of North America tallying 173 companies during last year’s annual meeting alone. And imaging experts must familiarize themselves with these tools to stay ahead of the game.
Radiologists with Augusta and Emory Universities in Atlanta set out to do just that with a special report published Nov. 11 in Radiology: Artificial Intelligence.
“The promise and potential benefits of radiology AI software continue to grow, and radiologists, practice administrators, and IT staff must continue to educate themselves on the potential benefits, drawbacks, and costs of implementation,” Yasasvi Tadavarthi, with Augusta’s radiology department, and colleagues wrote.
They analyzed 119 software offerings from 55 companies, including 46 with U.S. Food and Drug Administration and/or CE Mark approval as of 2019.
Radiology’s AI market is just getting started
“In comparison to other medical software, the radiology AI software market is in its infancy,” Tadavarthi and colleagues explained on Wednesday.
Such companies first made their mark at major imaging conferences— such as SIIM and RSNA—in 2016. Since then, vendors have been multiplying, with the world market cap for image analysis software expected to hit $2 billion by 2022, up from an estimated $1.2 billion in 2019.
Image interpretation companies can be broken down into four major task categories: repetitive, quantitative, explorative, and diagnostic. The latter makes up nearly half of all offerings, focusing on spotting pneumonia on x-rays and classifying breast lesions, among other things.
Many anatomic-specific tools are available, with untapped opportunities
A great number of vendors are focusing on lung CT and two-dimensional mammography offerings. The former has been popular due to the availability of large public datasets, while breast imaging’s immense exam volumes and high costs are drawing investor dollars.
On the flip side, there are fewer tools focused on cardiac and body imaging, including the abdomen and pelvic region. There are even less, however, dedicated to nuclear medicine, ultrasound, and pediatrics.
“These areas are ripe for innovation: Some potential high-volume targets include pediatric head US for neonatal hemorrhage, extremity US for deep vein thrombosis, or renal US for quantified grading of hydronephrosis,” the authors wrote.
Product purchasing considerations and pitfalls
As with any big purchase, one of the first questions should be "why am I buying this?" There are many answers to consider: quality improvement, efficiency gains, but also downsides such as costs and workflow changes.
Implementing AI also has its own pitfalls, the authors noted. Purchasers should investigate cost structures such as fee-for-service or subscription models, along with intangible costs like resistance to embrace new technologies.
Data ownership and patient privacy should also be considered before implementing new software, Tadavarthi et al. wrote. Federal regulations, data storage, and deidentification must all be top of mind.
Finally, the authors underscored the importance of having input from key stakeholders, such as administrators, clinicians and IT staff, when sorting through software options. Ensuring everyone is in agreement is key to a successful future, they noted.
“Deployed correctly, these [tools] can be a boon to both patients and providers in an ever-evolving health care setting with increasing imaging volumes and complexity,” the group wrote.