Study of 3D morphological filtering applied for automatic detection of lung cancer X-ray CT

In this paper, we improved our automatic lung cancer detecting algorithm for computer aided diagnosis system in Lung Cancer Screening CT (LSCT). We had earlier developed a two-dimentional Quoit filter, which successfully detected lung cancers. But many false positive detections, basically blood vessels, were also observed. To further improve the effectiveness of our algorithm, we extended the Quoit filter, from two-dimension to three-dimension, and also developed a new method called MIP-2D-Quoit processing. These methods were also to reduce many false positive detections. For 7 samples, i.e. a total of 139 slice cross sections, the new method was able to reduce the total number of lung cancer candidates to only 20 slices. (author)