AUTOMATED SELECTION OF REGIONS OF INTEREST FOR QUANTITATIVE-ANALYSIS OF LUNG TEXTURES IN DIGITAL CHEST RADIOGRAPHS

In order to implement a computerized scheme for quantitative analysis of interstitial lung disease in chest radiographs in clinical situatio ns, a fully automated method of selecting many square regions of inter est (ROIs) in peripheral lung areas are developed. First, the peripher al lung regions are identified, based on the automated detection of lu ng apices, ribcage edges, and diaphragm. Then a large number of ROIs a re selected sequentially by filling in the identified peripheral regio ns. Finally, those ROIs containing sharp edges are removed based on an edge gradient analysis, for which a gradient-weighted edge orientatio n histogram is employed. Approximately 200-400 ROIs were automatically selected for each case with this method. The evaluation of using ROC analysis indicated that the automated ROI selection method was effecti ve in quantitative analysis of lung textures in digital chest radiogra phs.