Semiautomatic evaluation procedures for quantitative CT of the lung

Accuracy, reproducibility, and objectivity are important in quantitative assessment of lung density and structure by CT, and the measurement has to be carried out under tightly controlled conditions. We therefore employ respiratory gating at defined levels of inspiration for CT scanning. In the evaluation process, we found operator-induced errors of 2-6 HU standard deviation for the relatively simple task of global lung density estimates; in regional evaluations, they frequently exceeded 10 HU. We therefore developed semi-automated evaluation algorithms that isolate lung parenchyma by fast contour tracking and define subregions by shrinking, radial segmenting, and anteroposterior subdividing of the left and the right lung. Global and regional mean density values and histogram parameters were extracted. Based on our clinical studies, we estimate that an overall precision of better than 5% can be achieved in quantitative CT of the lung with cooperative patients.