Sign In

Communications of the ACM

ACM TechNews

Machines Rival Expert Analysis of Stored Red Blood Cell Quality


View as: Print Mobile App Share: Send by email Share on reddit Share on StumbleUpon Share on Hacker News Share on Tweeter Share on Facebook
An illustration of human red blood cells.

New artificial intelligence models can automate evaluations of stored red blood cells.

Credit: Sebastian Kaulitzki/Getty Imahges

A three-year, multi-institutional study by 19 researchers at 12 institutions in five countries has yielded artificial intelligence (AI) models that can automate evaluations of stored red blood cell (RBC) quality that match or surpass expert assessment.

The investigators used 40,900 cell images to teach neural networks to classify RBCs into six categories; a fully-supervised machine learning algorithm agreed with human experts’ assessments 77% of the time (even human experts only agree 83% of the time).

Ryerson University's Michael Kolios calls this achievement "a testament to how technology and science are now interconnecting to solve today's biomedical problems."

From Ryerson University News
View Full Article

 

Abstracts Copyright © 2020 SmithBucklin, Washington, DC, USA


 

No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account