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

Interactive multimodal breast cancer diagnosis based on a registration of X-ray mammograms and 3D volume data

Meeting Abstract

Suche in Medline nach

  • Torsten Hopp - Karlsruhe Institute of Technology, Karlsruhe, DE
  • Nicole Ruiter - Karlsruhe Institute of Technology, Karlsruhe, 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.326

doi: 10.3205/13gmds259, urn:nbn:de:0183-13gmds2597

Veröffentlicht: 27. August 2013

© 2013 Hopp 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: Combining complementary multimodal information is likely to benefit early breast cancer diagnosis. Yet, the current screening method X-ray mammography images the breast in a considerably different configuration (2D projection, breast compression) than upcoming 3D modalities like Magnetic Resonance Imaging (MRI) or Ultrasound Computer Tomography (USCT). To overcome these differences, a 2D/3D model-based image registration has been developed, which is able to predict the spatial relationship between 2D mammograms and arbitrary 3D modalities [1], [2]. Intuitive visualization of the registration results is the key to gain maximal profit from multimodal information for radiologists. In this work we present results of our visualization techniques and our interactive viewer software for multimodal diagnosis.

Methods: Mammograms and 3D images are registered using our fully-automated image registration method, which is based on simulating the mammographic compression of the breast by a biomechanical Finite Element Model [1]. Registered images have overlapping circumferences and can be compared directly. To visualize the relationship between the modalities, two techniques have been implemented: image fusion and interactive point tracking. Image fusion is carried out by extracting quantitative information from the 3D modality, e.g. the contrast enhancement in MRI [3] or sound speed in USCT [4]. This information is color-coded and projected semi-transparently on top of the gray level mammogram. The presentation of the color-coding can be adjusted by the user. Interactive point tracking is based on the deformation field calculated during the image registration. For a point of interest selected by the user, the according position in the complementary modality is calculated and displayed. Using two-view mammograms, the 3D position of a point of interest can be estimated.

Results: The registration was applied to 79 clinical MRI and 13 USCT datasets. The registration accuracy was evaluated by the displacement of landmarks, which were annotated by experts in the 3D images as well as in the mammograms. The mean registration accuracy was 13.3 mm for the MRI and 10.4 mm for the USCT images. Image fusion with MRI contrast enhancement was carried out and presented to radiologists. In 88% of the cases, additional information contributed by the fused mammograms against conventional mammograms was found. Image fusion with sound speed demonstrated that USCT images are likely to be able to distinguish predominant tissue types within the breast (fat, gland, and tumor) by absolute sound speed. Point tracking was used to predict the 3D position of a point of interest in the MRI images. The mean localization accuracy was 14.3 mm [5]. Both techniques have been integrated into a DICOM viewer software, allowing interactive assessment.

Discussion: The evaluation of our fully-automated registration method with clinical images obtained promising results in a clinically valuable range. The registration serves as a basis to create interactive visualization of the complex spatial relationship between modalities. Our integrated viewer software allows for intuitive access to the registration results. The first clinical trial on the image fusion of mammograms with quantitative data from the 3D modality was encouraging for future studies.


Hopp T, Dietzel M, Baltzer PA, Kreisel P, Kaiser WA, Gemmeke H, Ruiter NV. Automatic multimodal 2D/3D breast image registration using biomechanical FEM models and intensity-based optimization. Med Im Analysis. 2013;17(2):209-218.
Hopp T. Multimodal Registration of X-Ray Mammograms with 3D Volume Datasets [PhD Thesis]. University of Mannheim; 2012.
Hopp T, Dietzel M, Baltzer PA, Kaiser WA, Ruiter NV. 2D/3D Image Fusion of X-ray Mammograms with Breast MRI: Visualizing Dynamic Contrast Enhancement in Mammograms. Int J Comput Assist Radiol Surg. 2012;7(3): 339-348.
Hopp T, Duric N, Ruiter NV. Automatic multimodal 2D/3D image fusion of ultrasound computer tomography and x-ray mammography for breast cancer diagnosis. Proceedings SPIE Medical Imaging. 2012;8320.
Hopp T, Ruiter NV. 2D/3D Registration for Localization of Mammographically Depicted Lesions in Breast MRI. In: Proceedings 11th International Workshop on Breast Imaging, 2012, p. 627-634.