MIT’s Computer Science and Artificial Intelligence Lab has built up a new deep learning-based AI prediction model that can anticipate the development of breast cancer as long as five years ahead of time. Researchers working on the tool additionally recognized that other similar projects have often had an inherent bias since they were based on white patient populations, and explicitly planned their own model so it is informed by “more equitable” data that ensures it’s “equally accurate for white and dark women.”
That is critical, MIT notes in a blog post, since black woman is more than 42 percent more likely than a white woman to die from breast cancer, and one contributing element could be that they aren’t too served by currently early detection techniques. MIT says that its work in developing this technique was aimed specifically at making the assessment of health risks of this nature more accurate for minorities, who are often not well represented in the development of deep learning models. The issue of algorithmic bias is a focus of a lot of industry research and even more up to date products forthcoming from tech companies working on developing AI in the field.
This MIT tool, which is prepared on mammograms and patient outcomes (eventual development of cancer being the key one) from more than 60,000 patients (with more than 90,000 mammograms total) from the Massachusetts General Hospital, starts from the data and uses deep learning to identify patterns that would not be clear or even perceptible by human clinicians. Since it’s not based on existing presumptions or required knowledge about risk factors, which are at best scenario a suggestive framework, the results have so far demonstrated to be far more accurate, particularly at predictive, pre-diagnosis discovery.
Overall, the project is intended to help healthcare professionals put together the correct screening program for people in their care and eliminate the heart-breaking and all-too-common outcome of late diagnosis. MIT hopes the technique can likewise be used to improve detection of other diseases that have comparable issues with existing risk models with far too many gaps and lower degrees of accuracy.