Fully automatic detection and quantification of emphysema on thin section MD-CT of the chest by a new and dedicated software

Purpose: Introduction of a novel software tool (YACTA - yet another CT analyzer) for detection and quantification of pulmonary emphysema in thin-slice chest MDCT data sets. Materials and Methods: Consisting of grey-level threshold-based algorithms (e. g., region-growing), expert rules and morphological image postprocessing YACTA segments the tracheobronchial tree prior to the detection and quantification of pulmonary emphysema. In addition to general parameters, such as the mean lung density (MLD) and the emphysema index (EI - also described as pixel index PI), the previously described bullae index (BI) is transformed into a three-dimensional parameter for a morphological description of emphysema. A first evaluation of chest MDCT data sets of 11 patients was performed as well as a comparison of MLD, lung volume (LV), emphysema volume (EV) and PI calculated with two established commercial tools of Siemens Medical Solutions (Volume(C) and Pulmo(C)). Furthermore, the BI was calculated with YACTA. Results: YACTA processed the image data without manual interaction and demonstrated more user-comfort than (Volume(C) and Pulmo(C)) software, which require manual correction especially for lung segmentation at the hilar regions to separate central airways from lung parenchyma. MLD, LV, and EV values calculated with YACTA were systematically higher (Pulmo(C): + 50 HU/+ 597 ml/+ 159 ml; Volume(C): + 32 HU/+ 110ml/ + 155 ml). Different segmentation algorithms are responsible for this: YACTA includes areas not assessed by mere threshold-based techniques. Constantly lowered LV values of Pulmo(C) are caused by a missing dilatation algorithm. The error correction as a special feature of YACTA results in increased emphysema volumes and indices. The segmentation of the tracheobronchial tree lowers the part of airways falsely classified as emphysema. Conclusion: The new developed software shows higher user comfort as established by semi-automated tools. Results of LV, EV, MLD and PI are comparable or moderately different. Automatic calculation of a BI is possible, providing information about bullous morphology of pulmonary emphysema. Further studies are necessary to correlate data with clinical or pathological parameters.