gms | German Medical Science

GMDS 2013: 58. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e. V. (GMDS)

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie

01. - 05.09.2013, Lübeck

Lung Registration using Automatically Detected Landmarks

Meeting Abstract

  • Thomas Polzin - Universität zu Lübeck, Institute of Mathematics and Image Computing, Lübeck, DE
  • Jan Rühaak - Fraunhofer MEVIS Project Group Image Registration, Lübeck, DE
  • René Werner - Universitätsklinikum Hamburg-Eppendorf, Institut für Medizinische Informatik, Hamburg, DE
  • Heinz Handels - Universität zu Lübeck, Institut für Medizinische Inormatik, Lübeck, DE
  • Jan Modersitzki - Universität zu Lübeck, Institute of Mathematics and Image Computing, Lübeck, DE

GMDS 2013. 58. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e.V. (GMDS). Lübeck, 01.-05.09.2013. Düsseldorf: German Medical Science GMS Publishing House; 2013. DocAbstr.325

doi: 10.3205/13gmds258, urn:nbn:de:0183-13gmds2583

Veröffentlicht: 27. August 2013

© 2013 Polzin et al.
Dieser Artikel ist ein Open Access-Artikel und steht unter den Creative Commons Lizenzbedingungen ( Er darf vervielfältigt, verbreitet und öffentlich zugänglich gemacht werden, vorausgesetzt dass Autor und Quelle genannt werden.



Introduction: Lung diseases have a high prevalence and especially lung cancer is a common cause of death. A proper lung registration could be used in terms of follow-up-analysis, motion-correction during radiotherapy, treatment planning and diagnosis of e.g. emphysema or fibrosis. This paper proposes a novel approach for image registration of lung CT scans. A unique combination of automatically detected landmarks [4] and a method called CoLD [2] enables a landmark-based image registration without human interaction.

Methods and Materials: CoLD is deduced from ”Combining Landmarks and Distance Measures”. The pre-registration of the proposed approach is the landmark-based Thin-Plate-Spline (TPS) method. It is improved by an additional minimization of an objective function consisting of a Normalized Gradient Field distance measure and a curvature regularizer. As a special property, landmark correspondences established by the TPS registration are guaranteed to be preserved. This is done by optimization on the kernel of discretized landmark constraints. This kernel is computed with a singular value decomposition [3]. All experiments were performed on ten publicly available data sets provided by the DIR-Lab [1] which include the scans and 300 manually annotated landmarks that were used for evaluation of the methods.

Results: The CoLD technique showed superior results compared to the TPS registration incorporating the same landmarks provided by an algorithm specialized on landmark detection in pulmonary CT data. Integration of additional knowledge represented by automatically detected landmarks makes the nonlinear registration very robust and provides a magnificent starting point. Furthermore, CoLD revealed considerable advantages in comparison to an intensity-based registration using Normalized Gradient Fields and curvature regularization. In comparison with other recently published approaches our method achieves state of the art results concerning landmark error. Additionally, the magnitude of deformation is in a plausible range and no foldings of the grid occur.

Discussion: We conclude that the usage of automatically detected landmarks combined with a CoLD registration improves registration quality compared to both considered alternatives. An extension to other applications is easily accomplishable due to a modular implementation by an adjustment of distance measure, regularizer and landmark detection.


Castillo R, Castillo E, Guerra R, Johnson VE, McPhail T, Garg AK, et al. A Framework for Evalua- tion of Deformable Image Registration Spatial Accuracy using Large Landmark Point Sets. Physics in Medicine and Biology. 2009; 54(7):1849–1870.
Fischer B, Modersitzki J. Combining Landmark and Intensity Driven Registrations. Proceedings in Applied Mathematics and Mechanics. 2003;3(1): 32–35.
Haber E, Heldmann S, Modersitzki J. A Scale-Space Approach to Landmark Constrained Image Registra tion. In: Proceedings of the Second International Conference on Scale Space Methods and Variational Methods in Computer Vision (SSVM). Springer LNCS; 2009.
Werner R, Duscha C, Schmidt-Richberg A, Ehrhardt J, Handels H. Assessing Accuracy of Non-linear Registration in 4D Image Data using Automatically Detected Landmark Correspondences. In: SPIE Medical Imaging 2013: Image Processing. 2013.