Do not fear AI, embrace it
Despite AI experts predicting the replacement of radiologist with AI within the next few years, the radiologist’s future is not as dark as it seems.
Comments from many leading specialists on artificial intelligence (AI) have sparked concerns among radiologists and medical students considering a career in radiology. For example, Geoffrey Hinton, known as the godfather of deep learning, believes that in 5 years deep learning will be better at analyzing medical images than a radiologist. Others have questioned the need for radiologists in the near future.
What to make of these predictions? „Not all hype, not all reality“, Kim commented, “but a little bit of both”. AI is not all it is made up to be. AI algorithms misclassified road signs after black and white stickers had been put on them. In another instance, AI was not able to differentiate between a turtle and a rifle. And maybe the best-known example of AI gone wrong is Microsoft’s twitter bot that went racist within 24 hours. Kim added that many of the world’s smartest people have made predictions on the future that turned out wrong. “Take Bill gates for example”, Kim said. “In 2004, Gates predicted that the problem of spam would be solved within two years.“
The black box problem
Kim explained that AI is like a black box. “You do not know what the algorithm is doing”, he emphasized. IBM’s artificial intelligence program Watson was used to determine if patients should receive an MRI contrast agent. For one of the instances where Watson assigned a contrast agent in error, the confidence of the algorithm in its decision was ninety-nine percent. “Would you trust such a machine without any human interaction?” Kim inquired.
The bigger picture
The discussion on AI in radiology mainly focuses on the aspects of interpretation and reporting. Kim asked radiologists to “widen their field of view”. There are numerous other processes within radiology, where AI can have a positive impact on a radiologist daily routine, with scheduling and protocol optimization being just two examples. AI may also help overcoming problems experienced by radiologists today. In the United Kingdom and in the United States the numbers of radiologists diagnosed with burnout are high. AI may be part of the solution to reducing workload in radiology.
Radiologists embrace technology
“Radiologists have always been good at adopting new technologies”, Kim pointed out. He expects the same to hold true with regard to AI. “We should focus on figuring out how to use the positive potential of AI, rather than focusing on the negative aspects”, Kim recommended. Kim ended his talk with a quote from Curt Langlotz, a Professor of Radiology and Biomedical Informatics at Stanford University. While Langlotz does not believe that radiologists will be replaced, he does believe that “radiologists who use AI will replace those who don’t”.
It is not a question of radiologists or AI. Radiology will evolve and integrate the technological advances in AI into the radiological routine – for the benefit of the patients and the radiologists.