Thomas H. Davenport, PhD, a professor of management and IT at Wellesley, Mass.-based Babson College, and Keith J. Dreyer, DO, PhD, vice chairman of radiology and chief data science officer at Boston-based Massachusetts General Hospital, argued that while AI will change radiology, these innovations will support practicing radiologists, not replace them.
In recent years, researchers have developed AI systems that identify pathologies from medical images with relatively the same success as humans. By integrating AI into clinical practice, hospitals may be able to reduce human labor, lower costs and improve diagnostic accuracy, according to some researchers.
“We’re confident, however, that the great majority of radiologists will continue to have jobs in the decades to come — jobs that will be altered and enhanced by AI,” the authors wrote. “The productivity improvements may even mean that radiologists can spend more time doing what many of them find most fulfilling: consulting with other physicians about diagnoses and treatment strategies.”
Here are four reasons the authors say AI won’t replace radiologists.
1. Radiologists don’t just read and interpret medical images. They also consult with physicians on diagnosis and treatment, perform image-guided interventions and discuss procedures and results with patients, among other tasks.
2. While AI systems have been successful in interpreting radiology images in recent research, they are still far from ready for clinical use. Many systems are focused on a specific problem, such as the probability of a lesion or of cancer, rather than a comprehensive review of an image.
3. AI algorithms must be trained on labeled data, which, in radiology, comprises “images from patients who have received a definitive diagnosis of cancer, a broken bone or other pathology,” according to the authors. There’s no aggregated repository of radiology images in existence today, so collecting and labeling a database will be a time-consuming project.
4. Even once automated image analysis becomes possible on a larger scale, providers will need to wait for appropriate medical regulation and health insurance policies to incorporate the technology.
“Who’s responsible, for example, if a machine misdiagnoses a cancer case — the physician, the hospital, the imaging technology vendor or the data scientist who created the algorithm?” the authors wrote. “All these issues need to be worked out, and it’s unlikely that progress will happen as fast as deep learning research in the lab does.”
To access Drs. Davenport and Dreyer’s op-ed, click here.
More articles on artificial intelligence:
US News & World Report ranks top grad schools for AI
Mass General researchers develop AI technique to speed image reconstruction in radiology
Survey: 21% of healthcare employees concerned about job security with shift to AI
