Experts push for new generalist radiology AI models that move beyond single tasks, ease financial concerns
Experts are pushing for new generalist radiology artificial intelligence models that move beyond single tasks and consolidate image interpretation assistance into one total package.
Scientists made their case in an editorial published Tuesday by Radiology, noting that narrow AI solutions suffer from financial limitations such as unsustainable price scaling and market fragmentation. Generalist AI could address these and other clinical and operational challenges, producing comprehensive reports that reduce radiologist effort and “unlock new value propositions.”
Recent advancements such as foundational models—trained on diverse datasets and adaptable to a wide range of downstream tasks with minimal training—pave the way for this method.
“[Generalist radiology artificial intelligence] offers comprehensive decision support by delivering tailored recommendations to various stakeholders, such as radiologists, referring physicians and patients,” corresponding author Pranav Rajpurkar, PhD, a professor of biomedical informatics at Harvard, and colleagues wrote Sept. 10. “This approach enables [generalist radiology AI] to improve financial sustainability by reducing the need for multiple point solutions—narrowly scoped AI tools each designed to tackle a single imaging task or pathologic abnormality.”
AI devices are typically offered through a subscription model, charged on a per site, workstation, radiologist or study basis. But as these vendors incorporate an increasing number of solutions, costs will inevitably escalate, the authors noted. Some companies charge up to $100,000 per solution, though costs vary widely, with the number of narrow AI tools needed to match a radiologist likely much higher.
“At some point, health systems could be priced out of these tools and may find it cheaper to hire an additional radiologist,” the authors advised.
Other costs come into play beyond initial licensure, including infrastructure and hardware. Add in an escalating number of solutions from various vendors, and a radiology group also can expect potential malfunctions that necessitate upgrades, repairs and other overhead costs. Plus, these point solutions complicate budgeting, with practices unable to predict the release of new offerings.
“Market fragmentation resulting from a wide spectrum of vendors and tools also increases the complexity of financial decisions,” the authors noted. “Vendors have different pricing structures and service agreements, making it difficult to evaluate disparate offers. Fragmentation also reduces the ability of providers to negotiate bulk discounts, as they are more likely to engage with multiple vendors.”
Rajpurkar and co-authors’ hope is that the consolidation of image interpretation assistance into one package can minimize costs and open the market to radiology providers who cannot afford current offerings.
“By having the choice to pay for one comprehensive tool rather than debate merits of mix-and-match approaches using multiple point solutions, users could better compare costs of competing investments,” the authors noted. “[Generalist radiology AI’s] multitask capabilities will also increase its appeal to diverse stakeholders. These stakeholders may range from private groups seeking efficiency gains to academic centers seeking more value from tools that can identify incidental and opportunistic findings while driving downstream care.”
You can read the full commentary in the official journal of the Radiological Society of North America here. Others penning the piece included Sid Dogra, MD, a diagnostic radiology resident with the NYU Grossman School of Medicine; Xiaoman Zhang, PhD, a postdoctoral fellow at Harvard; and Ezequiel Silva III, MD, with South Texas Radiology in San Antonio.
Marty Stempniak has covered healthcare since 2012, with his byline appearing in the American Hospital Association's member magazine, Modern Healthcare and McKnight's. Prior to that, he wrote about village government and local business for his hometown newspaper in Oak Park, Illinois. He won a Peter Lisagor and Gold EXCEL awards in 2017 for his coverage of the opioid epidemic.