gms | German Medical Science

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

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie

15. bis 18.09.2008, Stuttgart

Dynamics of alveolar geometry obtained by automatic tracing of area changes within microscopy videos

Meeting Abstract

  • David Schwenninger - Universitätsklinikum Freiburg, Freiburg, Deutschland
  • Knut Möller - Hochschule Furtwangen, Furtwangen, Deutschland
  • Hui Liu - Universitätsklinikum Freiburg, Freiburg, Deutschland
  • Hanna Runck - Universitätsklinikum Freiburg, Freiburg, Deutschland
  • Josef Guttmann - Universitätsklinikum Freiburg, Freiburg, Deutschland

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie. 53. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie (gmds). Stuttgart, 15.-19.09.2008. Düsseldorf: German Medical Science GMS Publishing House; 2008. DocMI8-3

The electronic version of this article is the complete one and can be found online at:

Published: September 10, 2008

© 2008 Schwenninger et al.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License ( You are free: to Share – to copy, distribute and transmit the work, provided the original author and source are credited.




The recent development [1] of an endoscopic microscope allows to observe alveoli in-situ and in-vivo in animal models of lung diseases during mechanical ventilation.

Figure 1 [Fig. 1] illustrates the endoscopic microscopy as applied in animals during mechanical ventilation as well as a typical pressure and flow curve of one respiratory cycle.

Recorded video sequences of alveoli shall be analysed in relation to the applied respiratory pattern during mechanical ventilation. Determination of changes in alveolar area over time - done by hand - is time consuming, places a high cognitive load to the examiner and the reproducibility of the results is rather low [2].

This project aims to establish a computer-assisted tool that provides semi-automatic evaluation of video sequences of alveolar tissue acquired by means of alveolar endoscopy. The software tool contains routines to detect the alveolar borders and to calculate the alveolar area.


Section filtering (SF)

Since the results of linear image filtering (LIF) did not lead to satisfying results, a method to calculate the edges of an alveolus relative to its center point was developed.

Therefore every image segment is filtered separately with a rotated version of the specially designed LIF-kernel.

The result is an edge-map that does not include any edges in parallel to the tangent to the center point. Figure 2 [Fig. 2] shows a comparison between results of linear image filtering (LIF) and SF.


To detect the boundaries of an alveolus using the SF’s edge map, active contours, also known as “snakes” [3], have been used. Snakes are basically a number of ordered points that are moved on the image driven by an internal force and a counteracting external force. The internal force is calculated from the relative position of neighbouring points – and thus increases smoothness of the resulting curve. The external force is determined by the edge-map, thus driving the curve towards the edges.


To synchronize the recorded video with the respiratory data, which is recorded separately, the audio channel of the video-system was used. Amplitude modulation was used (cf. Figure 3 [Fig. 3]) to record low frequency - signals with a sound card. The input of the modulator was thereby connected to the supply voltage of the ventilators valves.


Figure 4 [Fig. 4] shows some results of a snake algorithm based on the resulting edge-map from the SF. The initialisation curve was placed close to the alveolar border. The snake algorithm adapts the curve to the border of the alveolus frame-by-frame.

Measured area size of different alveoli taken from the same video sequence is plotted in Figure 5 [Fig. 5].

As a measure of noise-reduction, several respiration cycles with identical ventilatory settings have been averaged. Figure 6 [Fig. 6] shows the resulting average cycle of the alveolar area during one breathing cycle of controlled mechanical ventilation.


The interpretation of a recognized alveolar border is not trivial, since the recorded videos only provide two-dimensional images of three-dimensional alveoli.

The videos reveal white “borders” around the alveoli which seem to mark their cross-sectional area in the plane.

If an alveolus would be assumed to behave like an elastic balloon, the area size of its cross section would be a reliable measure for the alveolar volume. The cross-sectional area of the alveoli correlates with respect to shape and frequency with the applied volume to a certain degree (Figure 6 [Fig. 6]). It appears as if the measured alveolar area highly correlates with the applied volume as long as the volume is low, and correlates less when the volume is high – in the latter case the alveolus expands orthogonal to the surface plane due to the mechanical interaction with adjacent alveoli which may limit alveolar expansion in the image plane.

After correlating the measured data of alveolar cross-sectional area with the respiratory data, a phase shift, if any, would still be a valid quantity. Another useful feature extractable from the videos might be the ratio between volume changes and alveolar area changes. It could allow new insights into the alveolar dynamics at low and high pulmonary gas volume.


This work is supported by grants of the MWK Baden-Württemberg,

DFG (Gu561/6-1) and Dräger medical, Lübeck.


Stahl CA, Schumann S, Knorpp H, Schneider M, Möller K, and Guttmann J. Intravital endo-microscopy of alveoli: a new method to visualize alveolar dynamics. J Biomech 39(Suppl 1): p598, 2006.
Schwenninger D, Möller K, Stahl C, Schumann S and Guttmann J. Alveolar microscopy: on the automatic determination of alveolar size during ventilation. Critical Care 2007, 11(Suppl 2), 2007
Kass M, Witkin A, and Terzopoulos D. Snakes: Active contour models. International Journal of Computer Vision 1: p321–331, 1987.
Tietze U and Schenk C. Halbleiter-Schaltungstechnik. Springer; 2002.