Artificial Intelligence: The Elephant in the Tumor Board Room
Academic Medicine, 2023
Joel Grunhut
Free full text | Read at the journal | View on PubMed
In brief
A letter describing a breast cancer tumor board where a cancer seen on new mammograms turned out to have been flagged three months earlier by newly installed AI software, which gave a 35% chance of malignancy. The team had set the prediction aside because no one could see what the AI saw. The case captures a question every clinical team will face: when to act on AI output that humans cannot verify, and whether physicians are trained to make that call.
Key findings
On the earlier mammogram, the AI outlined a suspicious area with a 35% malignancy prediction that the radiologists and the rest of the board could not appreciate, so it was dismissed.
The case left surgeons, pathologists, oncologists and radiologists facing a new ethical and decision-making problem: whether to act on AI predictions they cannot confirm with their own eyes.
In response, the author founded an AI interest group at his medical school, began a longitudinal research project on bringing AI into the curriculum, and advocated for national change in how physicians are prepared for AI.
How to cite
Grunhut J. Artificial Intelligence: The Elephant in the Tumor Board Room. Acad Med. 2023;98(5):540-542. doi:10.1097/ACM.0000000000005073