An article in Nature shows how an AI approach can explain racial disparities in the experience of pain that standard radiographic measures of the severity of pain couldn’t see.
Underserved populations experience higher levels of pain, for example, in the context of joint wear and tear. Doctors, using medical imagery, couldn’t pinpoint the reason for this higher pain. Therefore, they assumed it had to do with factors outside of the joint, like stress.
The researchers used deep learning to create a pain prediction model. It takes knee X-rays and predicts the level of pain that the patient will have. Interestingly, this predictive algorithm could explain way more of the racial disparities than the standard methods the doctors would use. This suggests that the extra pain these patients experience comes from within the knee, not from outside factors (as only the knee is visible in the X-ray).
The article concludes that using an algorithm like this could reduce the apparent racist impact of the current treatment decisions.
See: An algorithmic approach to reducing unexplained pain disparities in underserved populations in Nature.
Image from is partial screenshot for the original article in Nature.