Scientists are making great strides in artificial intelligence (AI) with each passing day. In the medical arena, AI holds the promise of future applications that improve medical practice in ways never before dreamed of. And yet, barriers remain. Clinical acceptance in radiology jobs provides the perfect illustration.
Studies have shown that AI’s predecessor, machine learning, can improve diagnostic accuracy. Moreover, the improved diagnostics can make radiology more accurate, more efficient, and more cost-effective. Yet radiologists themselves are not confident in the technology. Their lack of confidence appears to be slowing the adoption of machine learning across the specialty.
If clinicians are having trouble with machine learning, they are bound to have as much trouble with AI. Thus, some in the AI field are saying their biggest task right now is not improving machines. It is convincing clinicians that both machine learning and AI are good things for radiology.
Breast Cancer Screening with CAD
One of the more exciting areas of machine learning in radiology is observed in breast cancer screening. According to an analysis of nine peer-reviewed studies published in the JMIR Medical Informatics journal, computer aided detection (CAD) technology can be effectively used to interpret medical images, in place of the typical second reading by a trained pathologist.
CAD is a form of machine learning that scans images, analyzes data, and then uses that data to identify potential cancers in future scans. A Health IT Analytics report from mid-June 2019 said that use of the technology has been increasing over the last several years. Yet radiologists still are not confident in its abilities.
One particular study showed that 74% of surveyed radiologists agree with the idea of employing double reading to improve cancer detection rates. Yet only 55% agree that CAD technology can perform as well as a human pathologist.
Getting Radiologists on Board
It is understandable that radiologists would want a second reading performed by a peer rather than a machine. We humans innately trust one another because of our shared experience. Likewise, we are also skeptical of machines that reach conclusions based only on hard, cold data. Machines are incapable of intuition and reason. They are incapable of thinking outside the box.
Researchers say that the emphasis on machine learning and AI right now is focused mainly on technical capabilities. All of that is well and good, but technical capability means little if those who would most benefit from the technology are unwilling to use it. Therein is the difficulty so frequently observed in developing new medical technology.
The medical field is notorious for its slow adoption of technology. Even in something as benign as electronic record-keeping, the healthcare industry just slow-walks everything. The industry is naturally resistant to change. So it should be no surprise that, despite its technical effectiveness, CAD is not a big hit with radiologists.
The goal now is to put more emphasis on getting radiologists on board. Let the technology quietly develop in the background while AI evangelists focus on helping clinicians understand the benefits of using it. If they can get radiologists on board en masse, it will do more than anything else to improve the technical abilities of CAD.
Radiology Is Changing
Radiology, as both a medical specialty and healthcare field, is changing before our very eyes. Much of that change is the result of new technology. The only question that remains is how fast change will be implemented. If radiology is to include genuine AI at some point in the future, clinician acceptance is going to have to be positively addressed.