a radiologist might examine thousands of patient images a day
Shudder. As a young man, I used to inspect new cars for shipping damage. It was well known in the business that an inspector's effectiveness dropped off dramatically as the day progressed. Missed damage could cost our customer thousands of dollars, and yet even blatantly obvious stuff was missed from time to time. I once missed frame damage on a pickup truck that had been chained down a little too enthusiastically. The 45k truck was a write off.
Our inspections numbered in the "hundreds per day". I can't even begin to imagine the mental haze at the end of thousands.
Stating it as "thousands of patient images a day" is misleading. It would be the same as saying you inspected "thousands of parts each day". As the radiologist further down notes, CT scans contains many slices.
While computers don't get tired, they also have a really hard time solving stuff like annotation tasks automatically. One thing is getting a good enough general performance, another is to never make critical errors. I see a huge potential for ML approaches in health care, but primarily as an aid for the health care professionals and not as a full replacement.
Maybe not right now, but there is nothing to say that it can't eventually surpass humans in effectiveness and critical error rate. 10-15 years ago people would have said "yeah self-driving cars are good, but as an aid for the driver, never as a full replacement".
Yes, it's called decision fatigue and it's a real thing. Don't make important decisions in your life at the end of a long day where you've already made many, many decisions.
I don't think radiologists can examine and report on thousands of images. Have you seen a radiology report? It's a detailed 1-2 page writeup of findings. I am sure they have software that may autocomplete some stuff, or they dictate the findings to someone else. Still, the math doesn't check out.
Radiologist here. A chest abdomen pelvis cat scan I routinely read as about 500 axial images at 2.5 mm slices. Then there is a comparison or two or more. Then there are different phases of contrast for the liver, kidneys. Then there are coronal and Sagittal reconstructions. I easily look at an average of 2000 images per patient.
Voice recognition software autocomplete is about as good as your phone's autocorrect...
Shudder. As a young man, I used to inspect new cars for shipping damage. It was well known in the business that an inspector's effectiveness dropped off dramatically as the day progressed. Missed damage could cost our customer thousands of dollars, and yet even blatantly obvious stuff was missed from time to time. I once missed frame damage on a pickup truck that had been chained down a little too enthusiastically. The 45k truck was a write off.
Our inspections numbered in the "hundreds per day". I can't even begin to imagine the mental haze at the end of thousands.
I don't see Watson getting tired.