As artificial intelligence and machine learning are adopted across the industry, radiologists and pathologists are blasting the same alarm as countless others by asking whether AI-powered diagnostics are here for their jobs.
However, experts say that with the future of healthcare in augmented intelligence, not artificial, there is no reason to panic.
Companies like Verily, Microsoft and Philips have stepped into the new generation of pathology as old standards PreciseDx, Ibex and Paige have made headlines with new partnerships. Ibex recently partnered with AstraZeneca to develop an AI-based HER2 scoring product. PreciseDx received approval for AI-enabled breast cancer diagnosis from New York state.
PreciseDx Chief Medical Officer Michael Donovan, M.D., Ph.D., told Fierce Healthcare that despite AI playing a greater role in healthcare, humans still remain at the center of diagnostic technologies.
“I do believe it will make pathologists better and not in any way negate the field because at the end of the day, our regulatory and accredited systems require that a physician makes the diagnosis,” Donovan said. “With new tech, pathologists will be supported by tools that are standardized and quantifiable, and they can incorporate that into this entire process of diagnostic accuracy.”
PreciseDx focuses on the grading stage of the pathologist’s work. The company has honed in on both breast and prostate cancers, because grading systems for those cancers have “come in and out of favor,” according to Donovan.
The company’s tech provides a quantifiable standard with proven success by accurately predicting early-stage breast cancer recurrence. By following 2,000 patients from Mount Sinai Hospital, a study found that PreciseDx’s platform improved risk stratification and prediction of recurrence over histological grade and clinical features while supplementing molecular genomic tests.
The platform fits into pathologists’ workflows by highlighting areas they should take another look at using augmented intelligence. Augmented intelligence offers suggestions to humans, unlike artificial intelligence, which makes anonymous decisions.
“Yes, it is the future because we need tools that will help to augment what we currently do,” Donovan said. “Look, diagnostics is a subjective science. The tech supports and enriches the pathologist, and I think it makes them better.”
Despite expert urgings, medical students are still wary of the future of service specialties like radiology and pathology. A recent survey of medical students found that 23% of the 532 asked said they would not consider pursuing a career in diagnostic radiology.
The survey data, published in Academic Radiology, showed that between 2017 and 2021, the percentage of students who believed job prospects in radiology to be limited increased from 50% to 71%.
With the number of U.S. pathologists decreasing by 18% between 2007 and 2017, Donovan has spoken out, hoping that students will reconsider the specialty already experiencing a shortage.
PreciseDx CEO Wayne Brinster told Fierce Healthcare that when he speaks with health systems, there isn’t hesitation to invest in augmented intelligence because the No. 1 goal of diagnosis is the correct one.
“There's a big need to be able to well differentiate between somebody who's in high risk versus somebody who's lower risk, to know what is the next step right after you remove the lump,” Brinster said. “Everybody wants to make sure that there's no errors there.”
And as tech companies promise better health outcomes with new products, researchers like Hien Van Nguyen at the University of Houston are imagining how humans will be a part of the next generation of diagnostics.
Nguyen designates his new algorithm as computer-aided diagnosis. With a nearly million-dollar grant from the National Cancer Institute, he is turning the robotic eye on lung cancer diagnostics.
The approach Nguyen employs passively monitors the gaze of radiologists when reading digital slides. Based on previous research of radiologists’ vision patterns when looking for cancer versus pneumonia versus the flu, the algorithm can suggest that a radiologist looking for lung cancer should examine an area they haven’t fully assessed yet.
“I think you'll see across the industry that humans are still at the center,” Nguyen said. “To have an algorithm make a final diagnosis, you need regulation to support that. And who is responsible if that diagnosis is wrong? So that's a lot of hoops to go over until we reach the point where we completely trust a computer to make a prediction between life and death.”
When asked about technology readiness and if and when a day will come when technology replaces radiologists and pathologists, Hein says it’s too far in the future to be certain. Some people are saying 50 years, some people say 100. Anything, at this point, is a wild guess, Nguyen said.
“But as long as we want AI to serve our interests, then it should be human-centric. Drug discovery, cancer diagnosis, fighting climate change, those will never be non-human centric," he said.