Two methods for isolating the lung area of a CT scan for density information.

Abstract

Extracting density information from irregularly shaped tissue areas of CT [computed tomography] scans required automated methods when many scans were involved. Two computer methods are described that automatically isolate the [dog] lung area of a CT scan. Each starts from a single, operator-specified point in the lung. The 1st method followed the steep density gradient boundary between lung and adjacent tissues; this tracking method was useful for estimating the overall density and total area of the lung in a scan, because all pixels within the lung area are available for statistical sampling. The 2nd method found all contiguous pixels of lung that were within the CT number range of air to water and were not a part of strong density gradient edges; this method was useful for estimating density and area of the lung parenchyma. Structures within the lung area that were surrounded by strong density gradient edges, such as large blood vessels, airways and nodules, were excluded from the lung sample; lung areas with diffuse borders, such as an area of mild or moderate edema, are retained. Both methods were tested on scans from an animal model of pulmonary edema and were effective in isolating normal and diseased lungs. These methods are also suitable for isolating other organ areas of CT scans that are bounded by density gradient edges.