Radiology Societies Draft Statement on Ethics of Artificial Intelligence

Feb. 27, 2019
“AI has noticeably altered our perception of radiology data,” draft statement notes

As the pace of innovation involving artificial intelligence and imaging has increased, so has the level of concern about ethical issues surrounding AI’s use in radiology. A group of European and North American radiology societies is seeking comment on a draft statement describing potential ethical pitfalls.

“AI has noticeably altered our perception of radiology data --- their value, how to use them, and how they may be misused,” the statement says. “Rather than simply understanding AI, radiologists have a moral duty both to understand their data, and to use the data they collect to improve the common good, extract more information about patients and their diseases, and improve the practice of radiology.”

The statement is a joint effort by the American College of Radiology, European Society of Radiology, Radiology Society of North America, Society for Imaging Informatics in Medicine, European Society of Medical Imaging Informatics, Canadian Association of Radiologists, and American Association of Physicists in Medicine. Based on comments received, the group expects to release a final statement approximately six months after the close of the comment period on April 7, 2019.

Acknowledging that AI promises to improve quality, patient outcomes, and efficiency, and to decrease costs, the societies add that it will also bring new consequences, risks, and questions for both patients and the radiology community. The societies argue that radiology should start now to develop codes of ethics and practice for AI. Establishing regulations, standards, and codes of conduct to produce ethical AI will need to balance technical, clinical, and commercial motivations with appropriate moral concern, they write. “Ensuring ethical AI requires a desire to gain trust from all involved. Both radiology-centric AI expertise and ethical technology are needed to verify and validate AI products.”

Admitting that this topic will change rapidly as technology and data science advances, the authors describe this statement as suggesting an approach to ethics. Regulatory and legal frameworks are changing rapidly, too. California’s new data privacy law and the European Union’s GDPR and proposed Civil Law Rules on Robotics are harbingers of such legislation. “People who build commercial and generalizable radiology AI tools need instructive ethical guidance; this statement will help inform future groups charged with composing such regulations.”

The report is broken down into the ethics of data, the ethics of algorithms, and the ethics of practice. Regarding algorithms, the draft statement notes that “some types of AI commonly used in radiology, such as artificial neural networks, are “black boxes,” and historically it has been problematic to understand why they make specific decisions. This black-box approach is unacceptable for patient care, where decisions potentially have high consequences.”