Fast and Accurate Lung Image Registration

Clinical Challenge


In the diagnosis and therapy of lung diseases, non-rigid image registration methods yield new approaches. The registration of inhale and exhale scans provides insight into the progress of chronic obstructive pulmonary disease (COPD) by determining functional parameters. In lung cancer therapy baseline scans of the lung are compared to follow-up scans, which allows a temporal analysis on tumors, and can be used to evaluate the curing success. Image registration allows a quantitative evaluation of the therapy process.


Solutions and Methods


Fraunhofer MEVIS developed a tailored solution for lung registration, with sub-millimeter accuracy at a runtime of only 25 seconds. After being integrated into a research workstation developed by Fraunhofer MEVIS and the Diagnostic Image Analysis Group, Nijmegen, it has already been applied successfully to more that 10 000 low dose CT lung scans. Our algorithm is based on a solid mathematical framework and sound modeling. This enables us to guarantee reasonable results, for example by preventing tissue folding and unwanted volume changes. Moreover this enables us to easily adapt the method to other registration applications.



  • The fastest registration algorithm in MICCAI grand challenge EMPIRE10
  • Accuracy ranked 5th out of 37 participants
  • Robust registration algorithm already used on more than 10 000 lung cancer screening CT scans