Communications of the ACM
The First AI Breast Cancer Sleuth That Shows Its Work
The new artificial intelligence platform not only tells doctors where it’s looking, but which past experiences it is using to draw its conclusions.
Credit: A.J. Barnett et al
An artificial intelligence (AI) platform developed by researchers at Duke University analyzes mammography scans to determine whether patients should undergo an invasive biopsy and features an interpretable algorithm that shows how it made its decision.
The AI was trained with 1,136 mammographic images from 484 patients.
Initially, the AI was trained to locate suspicious lesions while ignoring healthy tissue and other irrelevant data, after which radiologists labeled the images to teach the AI to focus on the edges of lesions.
Duke's Alina Barnett said, "Other AIs are not trying to imitate radiologists; they're coming up with their own methods for answering the question that are often not helpful or, in some cases, depend on flawed reasoning processes."
The researchers found that the AI performed as well as other black box computer models, though it did not outperform human radiologists.
From Duke University Pratt School of Engineering
View Full Article
Abstracts Copyright © 2022 SmithBucklin, Washington, DC, USA
No entries found