A framework for the decomposition and features extraction from lung DICOM images

Extracting morphological features from DICOM images is useful to obtain numerical anatomic values for population-wide studies. Currently software tools on medical devices are able to extract some parameters that can indicate the presence of diseases. Nevertheless, there still is a lot of not exploited information contained in images which can be useful for research as well as to characterize human behavior. For instance, measures for lung volume compared with reference data sets can be studied starting from clinical images.

In this paper we report preliminary results on a framework for the acquisition and decomposition of DICOM images applied on a dataset containing lung exams from which we extracted information and parameters useful for disease research studies. The here proposed algorithms for images segmentation and anatomical features extraction have been tested on a clinical dataset obtained from University Hospital of Catanzaro, providing the framework validity.