AI Isn't Killing Radiology: It's Reshaping the Field

In 2017, Ezekiel Emanuel, an acclaimed oncologist and health policy analyst, stated that radiologists would will soon lose their jobs thanks to machine learning.

That hasn’t happened However, even though artificial intelligence hasn't replaced radiologists, it has dramatically transformed their field.

Over 75% of the AI software approved by the Food and Drug Administration for healthcare applications is aimed at aiding radiology practices, according to Curtis Langlotz, a radiology professor at Stanford University and president of the Radiological Society of North America’s board of directors.

"Radiology is at the forefront of incorporating AI into clinical practices," he notes. However, AI is not diminishing the requirement for human involvement.

Langlotz states, "AI isn’t a superior form of intelligence; it's merely a distinct type of intelligence. A combination of a human and a machine outperforms each on their own. This has held true from when I started researching AI in the '80s up until now."

Speed in urgent cases

About two-thirds of radiology departments in the United States use AI in some capacity, according to a recent unpublished survey from the American College of Radiology (ACR) Data Science Institute. The number has roughly doubled since 2019, says Christoph Wald, vice chair of the ACR’s board of chancellors and chair of its informatics commission.

Wald, who serves as a senior associate consultant radiologist at the Mayo Clinic alongside his work, mentioned that up until now, around 340 AI-powered radiological devices have received FDA approval, with this count continuing to grow. According to Langlotz, most of these tools consist of detection algorithms designed to identify various conditions ranging from brain tumors and pneumonia to breast cancer and stroke.

A CT scan of the body consists of numerous images that radiologists need to examine. AI tools can sift through this multitude of pictures to determine which ones are most probable to contain anomalies. A study A study published in the academic journal Neuroradiology revealed that AI tools can efficiently notify radiologists about crucial findings in head CT scans (like hemorrhages and hydrocephalus), enabling them to prioritize these cases.

“This enables us to prioritize these items and analyze them more quickly,” explains Langlotz, who additionally oversees the program. AI in Medicine and Imaging Center At Stanford University, they noted that this could positively impact critical scenarios, such as patients in the emergency room or ICU, allowing them to receive prompt attention since their imaging would be analyzed faster.

AI likewise holds the capability to provide patients with more precise outcomes.

Suppose I am considered an expert, perhaps even the top professional globally," states Elliot Fishman, a radiologist affiliated with Johns Hopkins Medicine, specializing in early pancreatic cancer detection through AI technologies. "AI might provide some assistance, yet not significantly." However, he adds, "If we consider your typical radiologist—most professionals fall into this category—the application of AI transforms them into experts. So, who gains from such advancements? Ultimately, it’s the patients who benefit.

Research has shown When two radiologists examine the same study, they may have differing opinions ranging from 3 percent to 5 percent of the time.

Pranav Rajpurkar, who serves as an assistant professor of biomedical informatics at Harvard Medical School and also co-founded a firm named a2z Radiology AI, looks forward to how AI can provide "another level of protection" by offering physicians a secondary review on all cases.

What research tells us

A randomized, controlled, population-based 2023 study published in the journal Lancet Oncology points to the promise of AI in radiology. More than 80,000 women in Sweden were randomly assigned either two radiologists to read their mammogram or one radiologist plus AI. The study determined that there was a similar cancer detection rate for both groups.

Another 2023 study , in the journal Radiology, found that one AI tool was very effective at ruling out abnormalities on chest X-rays, with a sensitivity of 99.1 percent. And one 2022 study A study published in the journal Frontiers in Public Health demonstrated that AI effectively identified lung nodules in CT scans.

However, AI tools are not infallible. 2024 study A study published in Radiology examined AI's capability to rule out specific conditions in chest X-rays. Despite AI demonstrating a high degree of precision in exclusion, errors made by the technology could have more severe clinical implications compared to those overlooked by a radiologist.

Many artificial intelligence detection tools generate false alarms which radiologists have to investigate further. According to Langlotz, "Artificial Intelligence designed to spot irregularities might end up generating additional tasks for radiologists." He suggests that this issue has hindered the implementation of certain AI algorithms.

Better reports, legal issues

Numerous emerging AI tools aim to assist radiologists with the arduous job of composing reports, potentially resulting in "substantial time savings" and aiding in the creation of "more accurate, higher-quality, and more uniform reports," according to Langlotz.

However, the implementation of AI in radiology also introduces intricate legal issues. In the U.S., responsibility still falls on the radiologist rather than the AI tech firm, as per Wald, since the AI-generated insights require endorsement from a certified doctor.

This situation might shift if an autonomous AI system operates without physician supervision within the U.S. marketplace. A self-governing radiological tool designed to interpret chest X-rays, developed by a firm named Oxipit, has received approval for utilization in the EU. Additionally, several American enterprises are currently exploring this idea.

Ahead of the curve

Even though artificial intelligence is currently trendy, it has long been integrated into the field of radiology, according to Despina Kontos, a computer scientist and professor of radiological sciences at Columbia University Vagelos College of Physicians and Surgeons. This technology was previously referred to as computer-aided diagnosis, yet it fundamentally served the same purpose. "Compared to other branches of medicine, radiology has somewhat stayed ahead in adopting computers and AI," she explains.

Many professionals believe that the probable scenario is not the replacement of radiologists by artificial intelligence, but instead, they will utilize this technology in various methods to enhance their precision and ultimately expedite their work process.

I believe what we're witnessing are truly remarkable and transformative times," Fishman states. "I'm not suggesting that AI will supplant radiologists, but rather that AI will play an essential role in aiding radiologists with study interpretation for many years to come.