Despite a growing body of research showing the potential benefits of artificial intelligence (AI), it's easy to remain skeptical.
Why we aren't AI-champions just yet
There are several explanations for why AI’s threat to radiologists has not materialized. Likewise, there are good reasons why radiologists themselves have not been captivated by the technology and eager to adopt it in their practices.
Scientific evidence for AI effectiveness remains uncertain. While media attention focuses on studies that show great potential for AI, there have been just as many less favorable study results.
Deep learning remains vulnerable to sabotage, false data, or faulty associations.
Radiologists have more knowledge about the context of imaging results, such as the full patient history than AI algorithms.
So, is there any way to win over radiologists to the use of AI in healthcare?
Saving time is the key to radiologist support
Instead of focusing on AI’s diagnostic capabilities, supporters really should focus on its ability to save radiologists’ time. Like any other career, radiology has its share of what Dr. Amine Korchi calls “time-consuming, repetitive and mind-numbing” tasks.
One of those tasks, measuring leg length discrepancy (LLD) in children, is the subject of a new study that could nudge radiologists toward a more pro-AI mindset. (You can check out the full study online in Radiology.)
How much time are we talking?
Researchers from China’s Yantai University and Children’s Hospital of Philadelphia created a deep learning model that used data from the X-rays of 179 children.
The deep learning method measured pediatric leg lengths 96 times faster than subspecialty-trained pediatric radiologists, with a high correlation to the radiologists’ measurements.
Although we may not all be in that situation, what we can take from this study is AI's potential to help with monotonous, everyday tasks. Some advanced imaging software, like Nova3D+, uses artificial intelligence to enhance images and speed up your process.
Do you already use AI in your workflow? Share how in the comments!