14 April 2011

New computer science research: Early diagnosis of KOL

10.000 lungs run through a 3D scan. A sample of this volume is a rare gift to a computer scienctist!

Lauge Sørensen, postdoc at the Department of Computer Science at the University of Copenhagen, saw the potential and reused image samples which were originally collected for screening lung cancer.

Lauge used the material in an algorithm which "learned" to recognize patterns indicating a probablity of the smoker's lungs disease (KOL).

The algorithm has through machine learning obtained the ability to distinguish various patterns of the lung tissue and draw a probability map of areas indicating "bad" tissue.

 

Read the entire article in videnskab.dk (in Danish).