• AI
  • Molecular Imaging
  • CT
  • X-Ray
  • Ultrasound
  • MRI
  • Facility Management
  • Mammography

For Radiology’s Future, Look to its Past


Experts discuss whether machines will replace radiologists at RSNA 2016.

With the advent of speech recognition software and computer-aided diagnosis (CAD), the question of whether technology will soon replace radiologists has been swirling around for years.

Now, the introduction of deep learning and machine intelligence has re-ignited this discussion, and industry leaders are fervently discussing what might happen to the specialty in the next 20 years.

At RSNA 2016, informatics experts Eliot Siegel, MD, vice chair of imaging informatics at the University of Maryland School of Medicine, and Bradley Erickson, MD, radiology professor with a focus on computer-aided diagnosis at Mayo Clinic, discussed this controversy.

Surpassing the Radiologist
The past two decades have seen significant advancements in digital tools, Erickson said. If this pattern continues, which is expected, radiologists should be worried that machines will soon take away their jobs, Erickson said.

Mammography CAD has shown machines can be effective at identifying radiological findings the human eye can miss. As deep learning becomes more pervasive, machine’s abilities will only expand, he said.

“Deep learning isn’t biased or limited to human intuition,” he said. “It’s taught important signs. It takes broad images and figures out what’s important. Computers can find features that aren’t humanly perceptible, but they reflect important diagnoses.”

For example, deep learning can be instructed to pinpoint and consider genomic properties with reading studies and contributing to diagnoses.

Radiologist Maintains Control
If you want to see what radiology will look like in 20 years, look back over the previous 20, Siegel said. Speech recognition software and mammography CAD may have hit the scene in the late 1990s, but their improvement has been incremental, at best. The radiologist is still integrally involved in the diagnostic process.

“Mammography CAD has been around for 25 years and was introduced as something that would replace the mammographer,” he said. “According to a recent survey, 89% of mammographers do always use CAD, but do they change their opinions based on what it finds? Only 2% do routinely, and 36% do sometimes. Sixty-two percent report rarely or never changing.”

These technological tools simply haven’t risen to the levels anticipated, Siegel said. Medical images are often too complex for these aids. Extrapolating information from medical images is more challenging than doing so with more simplistic images, limiting the tools’ ability to be as effective as possible.

In addition, Siegel said, getting approval for machines to take a radiologist’s place is impractical from a regulatory perspective. The amount of testing and efforts necessary to secure clearance from the Food & Drug Administration for removing the provider and leaving diagnoses to a machine would be overwhelming. Even though the machines can successfully predict some diagnoses, the judgement needed for more complex, nuanced studies is only available with a radiologist.

What Will Most Likely Happen
Both points of view are likely extremes of the pendulum swing, they said. The next 20 years will bring changes, but they’ll be smaller than many in the industry fear.

As they improve, computers will perform quantitative imaging and biomarker measurements. They’ll also create structured reports that radiologists can review and approve, Siegel said. Dozens of applications in workflow, patient safety, communication, quality assessment, and follow-up opportunities will present themselves, he said, but radiologists will be harder to replace than currently anticipated.

For patients, Erickson added, the combination of radiologists and machines will be beneficial. Together, you’ll see more important findings that are subtle. And, as machines assume more responsibilities, you’ll have more time to devote to patient interaction, leading to greater patient satisfaction and improved patient experiences.

Ultimately, Siegel said, any radiology resident concerned about his or her future should breathe easily. Your job is secure.

“I reassure any radiology resident that contacts me to finish their residency. There will be more radiologists – not less – in 20 years,” he said. “The future is secure with machine learning computers that will be trusted friends. Deep learning machines might read preliminary reports, but they won’t be reading by themselves in the United States. Radiologists will still judge, explain, quality check, counsel, teach, and discover in images.”

Related Videos
Can Diffusion Microstructural Imaging Provide Insights into Long Covid Beyond Conventional MRI?
Assessing the Impact of Radiology Workforce Shortages in Rural Communities
Emerging MRI and PET Research Reveals Link Between Visceral Abdominal Fat and Early Signs of Alzheimer’s Disease
Reimbursement Challenges in Radiology: An Interview with Richard Heller, MD
Nina Kottler, MD, MS
Related Content
© 2024 MJH Life Sciences

All rights reserved.