Radiology departments’ adoption of machine learning accelerates
Machine learning is growing in importance in radiology departments, as a majority of top imaging executives expect to use it within the next three years.
However, as more healthcare organizations look to adopt it, questions are arising about the potential impact on providers and the vendors that supply it, according to results of recent research by Reaction Data, a research firm that released a report this month on machine learning in medical imaging.
The organization undertook the study in light of growing excitement about the technology, particularly at the annual meeting of the Radiological Society of North America the previous November, said Jeremy Bikman, founder and CEO of the company.
Large technology companies are investing heavily in AI, and “in the healthcare arena, major players such as Change Healthcare, Nuance, Hologic and many other healthcare specific bellwethers have launched major AI initiatives,” Bikman adds. The firm’s research sought to determine the current state of machine learning use among healthcare organizations and make predictions about future adoption. Of the 133 respondents to the research, the majority of respondents were directors of radiology or other imaging executives, with the vast majority working at acute-care hospitals.
Of the respondents, some 45 percent said machine technology was either important or very important to their organizations. Only 16 percent of respondents asked about the importance of machine learning said it just wasn’t that important to their organizations. However, most respondents indicated they weren’t familiar with machine learning. Only 25 percent said they were very or extremely familiar with the technology.
More than half of respondents say that their application of machine learning lies in the future, at least more than one year away. Some 27 percent say they expect to adopt machine learning in the next two years; another 25 percent say they don’t expect to use the approach within three years, but are expecting to do so. Only 16 percent say they have no plans to use machine learning.
“One of the most fascinating findings to come from the research is that there has been very little adoption by imaging centers,” the researchers wrote. “All of the adopters are hospitals. This is something we did not expect, as imaging centers are less risk-averse than are their hospital counterparts.”
In terms of expected use, more than two-thirds of respondents (68 percent) say they expect to use machine learning in breast imaging exams. Another 61 percent anticipate using machine learning to interpret lung imaging exams, while 58 percent expect to use it to assist radiologists in assessing chest imaging examinations. Bone (38 percent), cardiology (36 percent) and liver applications (34 percent) are also in providers’ future plans.
“It’s not a stretch to predict that most radiology studies over time will have their own unique AI algorithms; the ‘land rush’ to create and patent dozens of new algorithms is already under way,” Bikman says.
Breast imaging is currently far ahead of other scan types in using machine learning, respondents indicated.
While Reaction Data believes that the future for provider use of machine learning is bright, questions still remain about the financial viability for vendors offering the technology.
“A critical issue is how are vendors going to make money selling their AI,” the report says. “Cost pressures in radiology and in other areas are very real. Reimbursement rate trends don’t paint a rosy picture, so how are AI solution vendors going to justify the additional expense over the long haul?”
Answers to those questions will impact how vendors make money on AI, Reaction Data concludes. It’s unknown “whether AI solutions end up replacing, or radically altering, current imaging solutions like PACS.”
More information on the report can be found here.