Acyclic directed graphs for automatic image analysis of lung parenchymal geometry.

We have developed a simple method for encoding pulmonary alveolar geometry suitable for morphometric analysis and modeling. The approach involves the use of 8-bit video microscopic images of lung which are thresholded to produce binary images. Each alveolar wall was thinned to its midline to produce a skeletonized image. We present an algorithm for converting such an image to an acyclic directed graph. Such a representation is quite suitable for mathematical treatment, lending itself to problems involving simulation or modeling. We present as an example application, the measurement of dihedral angles in rat lung using conventional histological sections.