Mark Traill, M.D., a breast radiologist affiliated with the University of Michigan Health West in Wyoming, Mich, has been using artificial intelligence (AI) clinically in tens of thousands of breast imaging cases for the past five years. According to Dr. Traill, the benefits include improved confidence, speed, and accuracy.
However, in a recent interview at the Radiological Society of North America (RSNA) conference, Dr. Traill cautioned that a prudent assessment and evaluation of AI software is necessary to evaluate its advantages and shortcomings. He stated that people shouldn’t expect immediate results from AI, especially within the first week of using an AI algorithm.
Dr. Traill suggested that people should request a trial to test the technology, but also emphasized the need to spend time with the technology to understand its strengths and weaknesses. This, he said, can be done through internal validation, where the AI’s capabilities are tested and evaluated within an organization. “After you’ve been able to use it and be comfortable with it, then you can really push the gas pedal harder and see the advantages,” he said.
He also urged radiologists not to make judgments about the AI software’s functionality for the first few months. Dr. Traill emphasized the importance of ongoing monitoring of AI’s capabilities, particularly in areas like mammography.
“It’s great for that,” he said. “We are monitored all the time. They’re keeping track of all our false positives and our callbacks through MQSA (Mammography Quality Standards Act). You need to do that for the AI too and that exercise, again, will make transparent more details about how the algorithm is working, and enable you to build trust sooner.”
Dr. Traill, an assistant clinical professor at the Michigan State University College of Osteopathic Medicine, has also worked with AI in breast imaging.
About The Author
