Comparison of computer versus manual determination of pulmonary nodule volumes in CT scans

摘要: Accurate nodule volume estimation is necessary in order to estimate the clinically relevant growth rate or change in size over time. An automated volume-measuring algorithm was applied a set of pulmonary nodules that were documented by Lung Image Database Consortium (LIDC). The LIDC process model specifies that each scan assessed four experienced thoracic radiologists and that boundaries are be marked around the visible extent nodules for 3 mm larger. Nodules selected from database with following inclusion criteria: (a) they must have solid component on minimum three CT image slices (b) all radiologists. A total 113 met selection criterion with diameters ranging 3.59 32.68 (mean 9.37 mm, median 7.67 mm). centroid of each used as seed point algorithm. 95 (84.1%) were correctly segmented, but one considered not meeting first selection criterion method; for remaining ones, eight (7.1%) structurally too complex extensively attached 10 (8.8%) were considered properly segmented after simple visual inspection radiologist. Since specifications, as aforementioned, instruct include both sub-solid parts, method core capability segmenting tissues augmented take into account also parts. We ranked distances estimates radiologist-based median of values. 76.6% cases closer than at least values derived manual markings, which sign very good agreement the radiologists' markings.