gms | German Medical Science

57th Annual Meeting of the German Society of Neurosurgery
Joint Meeting with the Japanese Neurosurgical Society

German Society of Neurosurgery (DGNC)

11 - 14 May, Essen

Variational non-rigid image registration, denoising and edge detection of multimodal image data in a joint framework

Simultane, variationale und nonrigide Bildregistrierung, Entstörung und Strukturerkennung von multimodalen Bilddatensätzen

Meeting Abstract

  • corresponding author J.E. Scorzin - Neurochirurgische Universitätsklinik Bonn
  • B. Berkels - Institut für numerische Simulation, Universität Bonn
  • J. Han - Institut für Informatik, Universität Erlangen
  • J. Hornegger - Institut für Informatik, Universität Erlangen
  • M. Rumpf - Institut für numerische Simulation, Universität Bonn
  • H. Urbach - Radiologische Universitätsklinik Bonn, Abteilung für Neuroradiologie
  • K.L. Schaller - Neurochirurgische Universitätsklinik Bonn

Deutsche Gesellschaft für Neurochirurgie. Japanische Gesellschaft für Neurochirurgie. 57. Jahrestagung der Deutschen Gesellschaft für Neurochirurgie e.V. (DGNC), Joint Meeting mit der Japanischen Gesellschaft für Neurochirurgie. Essen, 11.-14.05.2006. Düsseldorf, Köln: German Medical Science; 2006. DocFR.08.09

The electronic version of this article is the complete one and can be found online at: http://www.egms.de/en/meetings/dgnc2006/06dgnc058.shtml

Published: May 8, 2006

© 2006 Scorzin et al.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by-nc-nd/3.0/deed.en). You are free: to Share – to copy, distribute and transmit the work, provided the original author and source are credited.


Outline

Text

Objective: Image registration may become an important technology not only in medical image analysis but in neurosurgical planning and assessment of intracranial pathoanatomy.

Many applications treat different image processing operations separately although the solutions of these problems have proved to be interdependent and may benefit with regard to reliability and time and cost effectiveness if treated in a simultaneous method. Therefore, we have attempted to develop a joint framework to detect edge features of different brain image modalities, to denoise and match the respective images with a non-parametric transformation simultaneously. The method’s reliability was investigated with different neuroimaging data sets.

Methods: The image segmentation and denoising is based on the Ambrosio-Tortorelli approximation of the classical Mumford-Shah functional using a phase field to represent the image edge set. The transformation of the edge set of one image is to match the edge set of the other. We performed a preliminary medical evaluation on 2D uni- and multimodal data sets. We compared 2D T1- and T2- weighted MRI sequences in axial orientation (Philips Gyroscan NT-Intera, 1.5 Tesla, Eindhoven, NL) derived from patients with intrinsic brain tumors. Moreover we applied registration of preoperative multimodal data sets (MRI and CT scans of the brain) to investigate efficiency of the algorithm. Visual control of the results particularly with regard to image and edge alignment was performed on a PC screen.

Results: The algorithm showed a reliable edge detection of the registered images, which was ensured in an applied multi-scale implementation scheme. Edge detection and correct alignment during registration was achieved unimodal but different image weighting, e.g. T1- and T2- weighted MRI sequences, where edges and geometry are presented in different intensities due to different consistencies of tissue. Also, multimodal application (2D cranial MRI and CCT scans) produced encouraging results with regard to image matching and computational time and effort.

Conclusions: The method presented for simultaneous edge detecting, image denoising and registration of multimodal neuroimaging showed high efficiency with regard to image alignment and computational time and effort. This may offer a future perspective for neurosurgical planning.