Chapter

Biomedical Image Registration

Volume 7359 of the series Lecture Notes in Computer Science pp 79-88

Inverse-Consistent Symmetric Free Form Deformation

  • Marc ModatAffiliated withCentre for Medical Imaging Computing, Department of Medical Physics, and Bioengineering, University College London
  • , M. Jorge CardosoAffiliated withCentre for Medical Imaging Computing, Department of Medical Physics, and Bioengineering, University College London
  • , Pankaj DagaAffiliated withCentre for Medical Imaging Computing, Department of Medical Physics, and Bioengineering, University College London
  • , David CashAffiliated withDementia Research Centre, Institute of Neurology, University College London
  • , Nick C. FoxAffiliated withDementia Research Centre, Institute of Neurology, University College London
  • , Sébastien OurselinAffiliated withCentre for Medical Imaging Computing, Department of Medical Physics, and Bioengineering, University College LondonDementia Research Centre, Institute of Neurology, University College London

Abstract

Bias in image registration has to be accounted for when performing morphometric studies. The presence of bias can lead to unrealistic power estimates and can have an adverse effect in group separation studies. Most image registration algorithms are formulated in an asymmetric fashion and the solution is biased towards the transformation direction. The popular free-form deformation algorithm has been shown to be a robust and accurate method for medical image registration. However, it suffers from the lack of symmetry which could potentially bias the result. This work presents a symmetric and inverse-consistent variant of the free form deformation.

We first assess the proposed framework in the context of segmentation-propagation. We also applied it to longitudinal images to assess regional volume change. In both evaluations, the symmetric algorithm outperformed a non-symmetric formulation of the free-form deformation.