“The wise are instructed by reason, average minds by experience, the stupid by necessity and the brute by instinct.”
I have this fear that we have worked ourselves into an efficient, well-educated, highly tuned Bugatti of a specialty, only to succumb to the (dare I say) stupidity that surrounds us. I am about to rant a bit. Please bear with me and sit back for it, or just move along. Nothing to see here.
I have been thinking a lot lately of AI and “machine learning” and all the things that are on the minds of anyone who is a radiologist and under the age of 60, and you know what I mean. The fallacy of this line of reasoning (machines can do that radiology stuff, no problem!) is that it relies on a reasonable, well-envisioned, and LOGICAL approach to problems.
Okay, everyone that works at a place where everyone is logical, reasonable, and not acting 60-70% of the time by “necessity” or instinct (see above quote), raise your hand. Okay, as I look about, I see no hands up. Allow me to provide an example.
We were taught in residency, and then in fellowship, and now almost ad infinitum I make a point of teaching the young ’uns that traumatic subarachnoid hemorrhage sits in a few sulci over the convexity, often with evidence of other trauma, and that IT DOES NOT NEED VASCULAR IMAGING. Vascular imaging would be POINTLESS. It would only find an incidental thing, or just be a waste of healthcare dollars. Aneurysmal subarachnoid hemorrhage, on the other hand, is in the basilar cisterns, more diffuse, and may or may not accompany trauma. Sudden onset, non-traumatic acute severe headache. You HAVE to image the vessels. It’s your sworn duty to the patient.
So (and I KNOW you’ve seen this recently, if not daily), the ED scan on an older patient who falls down the stairs with a pretty good head laceration shows trace subarachnoid over the convexity. In one or two sulci. Classic case. You pull the junior residents over to make the point.
“Traumatic hemorrhage, no vascular imaging,” you say, appearing omnipotent in your pronouncement. You feel good. Educating young minds. And, 5 minutes later, from the ED, yes, indeed, comes the request for the CT angiogram: “Rule out aneurysm.”
At first maybe you get a little righteous indignation, talk to the ED folks or maybe the neurosurgery folks, and get blue in the face and do the CTA. But after a few months, you just give up. You just do it. I guess the question is, do you still have that traumatic vs. aneurysmal discussion with the residents and fellows? Do you allow yourself to fall to the lowest common denominator? Dumb yourself down, too, with the rest of the herd, since it seems so rampant?
So, what will our friend AI do in this scenario? Personal opinion, it will lock into a loop—“This does not compute. You want a CT angiogram, but none is indicated.”—and fry all its circuits, like Yul Brynner in Westworld. End of AI as a problem.
Keep doing that good work. Mahalo.Back To Top
Phillips CD. Wet Read: Stupidity. Appl Radiol. 2018;47(4):34.
Dr. Phillips is a Professor of Radiology, Director of Head and Neck Imaging, at Weill Cornell Medical College, NewYork-Presbyterian Hospital, New York, NY. He is a member of the Applied Radiology Editorial Advisory Board.