Holger Frank Boehm MD, Presenter: Nothing to Disclose

Ulrike I. Attenberger MD, Abstract Co-Author: Nothing to Disclose

Christian Fink MD, Abstract Co-Author: Nothing to Disclose

Maximilian F. Reiser MD, Abstract Co-Author: Nothing to Disclose

Reliable and accurate methods for objective quantitative assessment of parenchymal alterations in the lung are necessary for diagnosis, treatment and follow-up of pulmonary diseases. The purpose of this study is to apply non-linear texture measures based on the Minkowski Functionals in 3D to Multi-Detector CT of normal and pathologic lung tissue, to differentiate between normal, emphysematic and fibrotic parenchyma, and to compare the diagnostic potential with algorithms based on tissue densitometry

275 cubic volumes-of-interest (VOI, edge-length 40 pixels) were obtained from MDCT chest-CT (isotropic voxel-size, edge-length 0.6mm) of 21 subjects with and without pathology (emphysema, fibrosis). All VOIs are visually consensus-classified by two radiologists. Texture-features based on the Minkowski-Functionals (MF) as well as on the CT-attenuation-values are determined. Classification-results of both approaches are assessed by Receiver-Operator-Characteristic and Discriminant-Analysis.

For the virtual biopsies from normal lung tissue, mean MF3D_SW was 2.7±8.2, for those with fibrotic [emphysematous] tissue the value was 5.0±4.4 [2.1±2.1]. In the densitometry-based analysis the subset with normal pulmonal tissue had a mean density of -829±42 HE, those with fibrous [emphysematous] tissue had a value -675±106 HE [-845±72 HE]. No significant correlation between tissue density and MF3D_SW was observed. ROC-Analysis resulted in an AUC of 0.87 for identification of pathologic tissue specimens using our novel algorithm; AUC for the densitometric method was 0.66.

Topological measures based on the Minkowski Functionals in 3D allow to quantify the texture characteristics of lung tissue depicted by MDCT. The detection of pathologic tissue appears feasible. In comparison with conventional densitometric methodologies the level of correct classification with our algorithm is significantly higher.   

Automatic topological analysis of pulmonary tissue by MF may be capable of enhancing density-based analysis and improve monitoring textural changes with progression of pulmonary disease.

Boehm, H, Attenberger, U, Fink, C, Reiser, M, Characterization of Normal and Pathologic Pulmonary Tissue by Topological Texture Analysis Applied to Isotropic Chest-CT Datasets.  Radiological Society of North America 2008 Scientific Assembly and Annual Meeting, February 18 - February 20, 2008 ,Chicago IL. http://archive.rsna.org/2008/6010482.html