High expectations for AI in imaging lag behind reality

Register now

Despite high expectations for the use of artificial intelligence in imaging, use of advanced computing technologies to assist radiologists is still in the early stages.

A recent report from KLAS Research finds that “talk about AI in imaging has been more common than actual adoption.” Despite that sobering assessment, some leading provider organizations and their vendor partners “have begun to roll out the technology or are making plans to do so.”

The KLAS report, researched through interviews with 81 healthcare organizations—primarily large integrated delivery networks—are an effort by the researchers to detail early imaging AI deployments and plans. The report also seeks to identify which vendors the health organizations see as leaders in providing AI capabilities.

Most striking in the KLAS research is that only 17 percent of healthcare organizations it contacted are either live or piloting AI projects for imaging initiatives. Another 30 percent say they have plans or are making plans to use the technology, while more than half—53 percent—say they have no current plans for using AI in imaging.

“Most organizations live with AI today in imaging today are beta testing new technology in limited settings, but have committed internal resources to further research and development,” says the report, written by Monique Rasband and Emily Paxman. “Others are actively monitoring the evolution of imaging AI, with almost one-third preparing for adoption.

“Those that are not live or do not have near-term plans to adopt AI in imaging say more research and time are needed to identify use cases and prove an ROI,” concludes the writers.

In terms of gauging timeframes for deployment, only 21 percent of responding delivery care networks say they expect to go live with AI imaging projects in the next 12 months. Another 41 percent see AI projects going live in one to two years, while 38 percent expect these initiatives to take more than three years.

Even though the time horizon for deploying AI in imaging is a few years away for most, “these organizations are requesting demos, discussing potential use cases with peers who are already live and assessing the AI needs of their radiologists,” KLAS concludes.

Adoption of AI is limited and still in testing mode, and “many organizations that are live with some form of imaging AI (using) the tools … within a single department … as a pilot to learn more about AI’s potential,” the report concludes. “Those who are not live today expect to follow a similar path i.e. lower initial adoption in a controlled, limited setting.”

However, hopes for AI imaging is high, despite the cautious implementation approach, KLAS researchers say. Survey respondents “express the long-term desire to see AI in imaging widely deployed and deeply adopted.”

The KLAS report also surveyed respondents on vendors in the AI imaging space. While IBM Watson Health was most frequently cited for its efforts in the AI imaging space, in general responding healthcare organizations say the vendor community is still early in efforts to provide solutions that consistently meet industry needs and expectations.

Generalizing on its past research into emerging technology, KLAS concludes that successful vendors in the AI imaging space will need the following traits:

  • Clear expectations of what outcomes will be achieved, and the steps that both the customer and vendor need to take to realize the outcome.
  • Proactive, strategic relationships that emphasize ongoing communication.
  • A central focus on training that is of high quality and continues over time.
  • Strong data governance, which was identified as a key by provider organizations that have an AI program in place.
For reprint and licensing requests for this article, click here.