What is so unusual about this? E.g. do not black skin colored athletes routinely outrun other athletes on short distances? The physique of Asians tends to be shorter and thinner than whites, right? Asian athletes are more adapt to martial arts than the typical white or black athletes.
Are not e.g. certain cancer types more prevalent in some population groups than in others like whites and Asians?
Does this not suggest most likely there are some physiological or skeletal differences between e.g. white and black skin colored or Asian humans? Would this not be reflected in CT scans or x-ray images?
Caveat: I have not yet read the preprint paper below! However, the researchers of the paper below have a long and comprehensive chapter on "Investigation of possible mechanisms of race detection" and a chapter on " Impact of image resolution and quality".
To debias the systems, as suggested by researchers, seems to be bad advice e.g. for my reasons given above! Skin color blindness maybe generally desirable in society etc., but it maybe dangerous and irresponsible when it comes to medicine!
What else? As usual the term "racial bias" (Why not systemic racism? I am being facetious) applied here to suggest something sinister is possibly going on has been!
"... Algorithms trained to diagnose medical images can recognize the patient’s race — but how?
What’s new: Researchers from Emory University, MIT, Purdue University, and other institutions found that deep learning systems trained to interpret x-rays and CT scans also were able to identify their subjects as Asian, Black, or White. ...
Behind the news: Racial bias has been documented in some medical AI systems"
Behind the news: Racial bias has been documented in some medical AI systems"
"... Standard deep learning models can be trained to predict race from medical images with high performance across multiple imaging modalities. Our findings hold under external validation conditions, as well as when models are optimized to perform clinically motivated tasks. We demonstrate this detection is not due to trivial proxies or imaging-related surrogate covariates for race, such as underlying disease distribution. Finally, we show that performance persists over all anatomical regions and frequency spectrum of the images suggesting that mitigation efforts will be challenging and demand further study. ...
However, our findings that AI can trivially predict self-reported race - even from corrupted, cropped, and noised medical images - in a setting where clinical experts cannot, creates an enormous risk for all model deployments in medical imaging: if an AI model secretly used its knowledge of self-reported race to misclassify all Black patients, radiologists would not be able to tell using the same data the model has access to. ...
These findings suggest that not only is racial identity trivially learned by AI models, but that it appears likely that it will be remarkably difficult to debias these systems. ..."
No comments:
Post a Comment