Eye-tracking reveals radiologists' reading habits when utilizing AI support
Radiologists guided by artificial intelligence-enabled visual cues are more likely to spot breast cancer on screening mammograms.
That’s according to new research published this week in RSNA’s flagship journal Radiology. Rather than test AI’s ability to detect malignant lesions on imaging, which has been the focus of numerous studies, researchers instead explored how it could help radiologists assess their own interpretation processes.
This was achieved by using a small camera-based tracking device placed in front of readers’ screens. The device deploys infrared lights to illuminate the radiologists’ eyes to identify patterns in how they analyze images with and without use of an AI decision support tool.
“By analyzing this data, we can determine which parts of the mammograms the radiologist [focuses] on, and for how long, providing valuable insights into their reading patterns,” explained the study’s joint first author Jessie J. J. Gommers, MSc, from the Department of medical imaging at Radboud University Medical Center in Nijmegen, Netherlands.
Experts used the device to analyze the reading habits of 12 radiologists interpreting 150 cases, 75 malignant and 75 benign. Rads read the exams with and without the help of an AI tool that highlighted areas suspicious for cancer and provided scores ranging from 0 to 100 based on level of suspicion. The tracking system analyzed readers’ visual focus based on their eye movements during their interpretations and took note of how long they were centered on specific areas.
When utilizing AI assistance, radiologists spent more time focusing on areas containing cancerous lesions; they also were more accurate with the help of AI, but did not spend any additional time reviewing exams with AI versus without it.
“Radiologists seemed to adjust their reading behavior based on the AI’s level of suspicion: when the AI gave a low score, it likely reassured radiologists, helping them move more quickly through clearly normal cases,” Gommers said. “Conversely, high AI scores prompted radiologists to take a second, more careful look, particularly in more challenging or subtle cases.”
The team expressed optimism for how AI can positively affect radiologists’ performance, while also acknowledging there is the risk that readers may become over reliant on AI assistance. However, when considering the numerous studies that have determined AI is as accurate as radiologists (sometimes more accurate), the group suggested that this risk is quite low.
Learn more about the study findings here.
In addition to her background in journalism, Hannah also has patient-facing experience in clinical settings, having spent more than 12 years working as a registered rad tech. She began covering the medical imaging industry for Innovate Healthcare in 2021.