AI in medical imaging has made significant inroads -- MRI images can be analyzed by the pixel, for example -- but radiologists should feel secure that they won't cede their jobs to the software.
That message from analyst Brian Edwards stems from his observations about the complementary nature of AI in medical imaging. Also, AI has the ability to improve surgical planning and perhaps prevent additional biopsies for patients.
"[AI] takes the pixels one by one, converts it to a number and then runs analysis just like it does with words [and] free text," said Edwards, who studies healthcare AI at Chilmark Research in Boston. "Is this the image of a tumor? Yes, no."
From there, oncologists and radiologists together review the AI results of images that possibly contain tumors. Confirmed cases can further use AI analysis to determine where the tumor masses exist and where healthy tissues lives, which helps determine more precisely what areas to cut out during surgery or focus on with other treatments such as chemotherapy, Edwards said.
AI in medical imaging may also make cancer detection a bit easier on patients. "Imaging technology has gotten to the point now where biopsies aren't always necessary, which is a huge benefit to the patients because biopsies are painful, invasive and often need to be redone several times," Edwards said.
However, AI will not shove aside radiologists any time soon, he added. One reason for his prediction is that typically AI does not perform medical diagnoses.
"It's mostly second-opinion type of work, confirming a radiologist opinion or running against a whole lot of images really fast to provide some sort of decision support," he said. "But radiologists are still the ones making all the decisions, and they're the ones accountable for the treatments provided."