For example, let's say I'm looking at a chest x-ray. There is a pneumonia at the left lung base and I am clever enough to notice it. 'Aha', I think, congratulating myself at making the diagnosis and figuring out why the patient is short of breath.
But, in this example, I stop looking closely at the X-ray after noticing the pneumonia, so I miss a pneumothorax at the right lung apex.
I have made a mistake radiologists call 'satisfaction of search'.
My 'search' for the patient's problem was 'satisfied' by finding the pneumonia, and because I am human and therefore fundamentally flawed, I stopped looking for a second clinically relevant diagnosis.
An AI module that detects a pneumothorax is not prone to this type of error. So it sees something I did not. But it doesn't see something that I can't see. I just didn't look.
https://www.npr.org/sections/health-shots/2013/02/11/1714096...
I'm skeptical to the claim that AI isn't prone to this sort of error, though. AI loves the easy answer.
Ah, now I have a name for it.
When I've chased a bug and fixed a problem I found that would cause the observed problem behavior, but haven't yet proven the behavior is corrected, I'm always careful to specify that "I fixed a problem, but I don't know if I fixed the problem". Seems similar: found and fixed a bug that could explain the issue, but that doesn't mean there's not another one that, independently, would also cause the same observed problem.
That is, the models spot pathologies that 99.9999% of rads would spot anyway if not overworked, tired, or in a hurry. But, addressing the implication of your question, the value is actually in spotting a pathology that 99.9999% of rads would never spot. In all my years developing medical imaging startups and software, I've never seen it happen.
I don't expect to see it in my lifetime.
I agree with almost everything you've said here.
Except 'not in my lifetime', because I plan on living for a very long time, and who knows what those computer nerds will come up with eventually ;-)
The difficulty is likely in making a good training dataset of labeled images with pathologies radiologists couldn't see. I imagine in some cases (like cancer), you may happen to have an earlier CT scan or X-ray from the patient where the pathology is not quite yet detectable to the human eye.
There are a lot of things in medicine that aren’t in literature, but are well-known among certain practitioners. I’m an anesthesiologist and practice in an area with a large African-American population. About 10-15% (rough guess) of people of West African descent will have a ridiculously strong salivary response to certain drugs (muscarinic agonists). As in, after one dose their mouths will be full of saliva in seconds. We don’t have East Africans for comparison, so I can’t say it’s a pan-Bantu thing, but I have seen it in a Nigerian who lived here. Not in the literature, but we all know it. I had a (EDIT: non-anesthesia) colleague ask me about a hypersecretory response from such a drug. I said, oh, was he black? Yes, how did I know? Because we give those drugs all the time and have eyes. It’s very rare to see in European-descended populations.
Rads can see a person's race. We look at them.
That's the reason rads never train to determine race from a chest x-ray.
BTW, models don't need to train that either. Because if it's important, it's recorded, along with a picture, in the guy's medical record.
I'd just like to gently suggest that determining someone's race from an X-Ray instead of, say, their photograph, is maybe not how we should be burning training cycles if we want to push medical imaging forward. Human radiologists had that figured out ages ago.
According to the article, human experts could not tell race from chest X-rays, while the AI could to do reliably. Further, it could still determine race when given an X-ray image passed through a low pass filter or high pass filter, showing that it's not relying solely on fine detail or large features to do so.
Firstly, that doesn't tell us whether there is bias in data. That tells us whether or not there is bias in their data.
Secondly, it tells us it can train to spot things that human rads do not train to spot. It tells us nothing at all about whether or not an AI can train to spot things a human rad also trains to spot, but can't.
Human rads don't train to spot race. Why? Because they don't need to do so. Human rads do train to spot pathologies in as early a stage as possible. I've never seen an AI spot one at an earlier stage than the best human rads can. But I have seen several AIs fail to spot pathologies that even human rads at the median could spot.
That's the state of play today. And it's likely to remain that way for a long, long time. Human rads will be needed to review this work not because they are human, but right now, it's because human rads are just better. At the top end, human rads are not only better, but are manifestly superior.
They aren't studying bias in data. They were studying bias in AI. The data used was 850,000 chest x-rays from 6 publically available datasets. They aren't studying whether this dataset differs from the general public or has some kind of racial bias; that's irrelevant to the study.
> it tells us it can train to spot things that human rads do not train to spot
You're kidding yourself if you think you could determine someone's race with 97%+ accuracy from a chest x-ray if only you trained at it. The study authors (who are themselves a mix of radiologists and computer scientists) claim that radiologists widely believe it to be nearly impossible to determine race from a chest x-ray. No one is ever going to try to train radiologists to distinguish race from chest x-rays, so you'll always be able to hold out hope that maybe humans could do it with enough training. But your hope is based on nothing; you don't have a shred of evidence that radiologists could ever do this.
> I've never seen an AI spot one at an earlier stage than the best human rads can.
According to the article, AIs aren't trained to do this, because we don't have datasets to train this. You need a dataset where the disease is correctly labeled despite the best radiologists not being able to see it in the x-ray. Trained with a good enough dataset, they'd be able to see things we miss.