gms | German Medical Science

GMDS 2012: 57. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e. V. (GMDS)

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie

16. - 20.09.2012, Braunschweig

Detection and compensation of relative motion in endoscopic heart valve recordings by circular Hough transform

Meeting Abstract

  • Stefan König - Fraunhofer-Institut für Integrierte Schaltungen IIS, Erlangen, Deutschland
  • Markus Kondruweit - Zentrum für Herzchirurgie des Universitätsklinikums Erlangen, Deutschland
  • Michael Weyand - Zentrum für Herzchirurgie des Universitätsklinikums Erlangen, Deutschland
  • Thomas Wittenberg - Fraunhofer-Institut für Integrierte Schaltungen IIS, Erlangen, Deutschland
  • Sven Friedl - Fraunhofer-Institut für Integrierte Schaltungen IIS, Erlangen, Deutschland

GMDS 2012. 57. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e.V. (GMDS). Braunschweig, 16.-20.09.2012. Düsseldorf: German Medical Science GMS Publishing House; 2012. Doc12gmds050

DOI: 10.3205/12gmds050, URN: urn:nbn:de:0183-12gmds0504

Published: September 13, 2012

© 2012 König 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

Introduction: Endoscopic video recordings are regularly used to analyze movements of heart valves [1], [2], [3]. Due to the recording setup, non-predictable relative motion between optics and valves leads to difficulties while interpreting and comparing results [4]. A detection and compensation of this relative motion is a necessary step prior further analysis. A major challenge here is to distinguish between relative motion and the valve movements. In this contribution, an approach for motion detection based on a circular Hough-transform is presented which exploits the given anatomical structure.

Material and Methods: For this study different mitral bioprostheses have been installed in an artificial flow generator and observed with an endoscopic high-speed video camera at 2000fps and a resolution of 256x256pixels. Eleven recordings with several opening intervals were generated, which suffer from relative motion of different intensity.

While recording the valves it was ensured that the annulus was completely present during the complete period. This fixed circular structure around the leaflets is not deformed during the systolic and diastolic phases and thus a constant anatomic landmark. A circular Hough-transform is used to detect the annulus throughout the image sequence. The epicenter currently has to be determined manually inside the first image. The correct position is not essential. Mainly it serves as a starting hint for detecting the closest best fitting circle. Successive images are using prior detected epicenters as starting points. Since the diameter of the valves is known, minimal and maximal radii can be defined easily by adding an offset. Due to the fixation of the bioprostheses, motion orthogonal to the image plane is impossible. Therefore, the radius of the annulus is constant during the sequence. Also, the image-to-image relative movement is limited by the high recording frequency. Considering these assumptions, false detections can be prevented by generating an average circle from the last and upcoming circles.

After the detection of the annular circle within the complete image sequence, the relative motion can be estimated by determining motion vectors from the epicenter coordinates.

Results: The presented approach has been tested with the available recordings. A visual rating about the compensation of relative motion was promising. For a quantitative evaluation, the relative motion of randomly chosen sequences has been determined manually by selecting landmarks in subsequent images. Additionally, a synthetic relative movement was added to single image frames by translating the image with random, but known vectors. The distance between the known translation vectors and the detected epicenter coordinates describes the quality of the motion compensation. With a mean synthetic translation of 25.8 pixels, the remaining mean relative motion after the Hough-based compensation can be minimized to 1.3 pixel.

Discussion: Using a circular Hough-transform to determine the annulus as a constant landmark in endoscopic video recordings of heart valves offers a fast and quite simple approach for the detection and compensation of relative motion. Whereas other methods fail due to missing distinguishing between relative motion and leaflet movements, this approach exploits the presence and form of the annulus and leads to sufficient results.


References

1.
Erasmi A, Sievers H, Scharfschwerdt M, et al. In vitro hydrodynamics, cusp-bending deformation, and root distensibility for different types of aortic valve-sparing operations: remodeling, sinus prosthesis, and reimplantation. J Thorac Cardiovasc Surg. 2005;130(4):1044-9. DOI: 10.1016/j.jtcvs.2005.06.005 External link
2.
Condurache A, Hahn T, Hofmann U, et al. Automatic measuring of quality criteria for heart valves. SPIE Medical Imaging. 2007;2007(6512);65122Q-1-65122Q-11.
3.
Kondruweit M, Wittenberg T, Friedl S, et al. Description of a novel ex-vivo imaging and investigation technique to record, analyze and visualize heart valve motion under physiological conditions. In: 39. Jahrestagung der DGTHG; 2010.
4.
Friedl S, Herdt E, König S, et al. Determination of Heart Valve Fluttering by Analyzing Pixel Frequency. In: Workshop Bildverarbeitung für die Medizin. 2012. p. 87-91.