Artificial Intelligence: The Elephant in the Tumor Board Room

Abstract

To the Editor: As I observed the tumor board, case after case of breast cancers were discussed and appropriate multidisciplinary management was planned for each individual patient. Then came one case that changed everything. Case #9 was a patient with 2 sets of mammograms pulled up to the big screen for all to see. The radiologist explained that the first set of images taken 3 months ago appeared completely normal and no suspect pathology could be found. The radiologist went on further to point out the evident pathology on the second set of images taken this week, thereby raising suspicion for cancer. Everyone’s eyes darted back and forth between the images, struggling to find some sort of hint of cancer in the first set of images. The radiologist continued, “This miss would be entirely acceptable if not for a recent advancement in our practice.” She mentioned that a few days before having our patient’s first images taken, we had implemented a new artificial intelligence (AI) software in our system. Although that software predicted a 35% chance of malignancy in the first images, we did not see it and waved it off. The radiologist pulled up an image with the AI prediction overlay, outlining the area of suspicious malignancy of which to the people in the board room could not appreciate. She said, “We dismissed AI’s prediction because we did not see what it saw.” The surgeons, pathologists, oncologists, and fellow radiologists all looked visibly disturbed by this case and the new decision landscape they faced. A look of confusion at facing a new complicated ethical world could be seen on the physicians’ faces. Everyone sat silently, understanding the difficulty the future would bring: to act on AI data or not. After a brief silence, they continued discussing the patient’s case and treatment. As I sat there on the side, I knew that this little conference room held a giant question. This patient’s case challenged me to think of novel ways to learn about AI and incorporate it into my professional identity by founding an AI interest group at my medical school, beginning a longitudinal research project of incorporating AI in the curriculum, and most importantly, advocating for change on a national level. These actions were my answers to the blaring question: When another AI dilemma arises, will we be equipped to address the elephant in the board room?

Ähnliche Arbeiten