Artificial intelligence (AI) is at the forefront of technology right now and is bringing with it innovation to every sector that harnesses its potential. In medicine, or specifically in radiology, AI will improve imaging-driven diagnoses and treatment. Radiology is an ideal application for AI because machine learning systems can distinguish patterns and irregularities in large collections of data. Learning from millions of images, the software can make diagnoses with speed and accuracy.
Recently, researchers at Google devised AI models to spot four types of findings on human chest x-rays. Their paper, which was published in the journal Nature, claims that the model family, which was created using thousands of images across data sets with high-quality labels, demonstrated “radiologist-level” performance in an independent review conducted by human experts. Prior to this, scientists at Northwestern Medicine collaborated with Google AI to create a model capable of detecting lung cancer from screening tests better than human radiologists with an average of eight years of experience.
Although algorithms are capable of detecting abnormalities with precision and detect some things that human eyes are unable to see, there are still unknowns regarding this technology. Dr. Eric M. Rohren, M.D., Ph.D., professor and chair of radiology at Baylor College of Medicine and radiology service line chief for Baylor St. Luke’s Medical Center, raises a few key questions regarding the value of an AI algorithm in medicine. He asks, “What is the value of (AI) in a health care system? Are you truly improving patient outcomes and patient care by introducing a particular algorithm into your practice?” Wisely, his questions caution us against the adoption of technology for the sake of technology.
How likely is it that AI will replace doctors? At this point, the answer is unlikely. The role of the radiologist, however, may change as the use of AI becomes more widely adopted. Eric Walser, M.D., chairman of radiology at The University of Texas Medical Branch (UTMB) at Galveston foresees radiologists becoming managers of data rather than the creators of the diagnoses. Yet others see AI as a tool that is best used in conjunction with physicians to improve decision making without actually replacing the physicians who interpret the imaging.
Rohren, the chair of radiology at Baylor, puts it well: “Machines and machine learning are very good at information handling, but they are very poor at making judgments based on that information. … The radiologist will continue to be critical.”