Fig. 1

Examples of low attenuation cluster analysis using either the threshold of a fixed CT density or that of the 15th, 25th, or 35th percentile of the CT density histogram. a and b Coronal sections of CT images show low attenuation areas (LAAs) based on thresholds of (a) fixed − 950 Hounsfield Units (HU) and (b) the 15th, 25th, and 35th percentiles of a histogram of CT values from all pixels in whole lungs. In 3-dimensional representation, the LAA clusters were expressed as spheres with volumes equivalent to those of actual clusters. Pink, yellow, and blue regions indicate large (> 500 mm³), moderate (> 50 and < 499 mm³), and small (< 50 mm³) clusters, respectively. Small clusters were omitted in the 3-dimensional representation. c The log-transformed volumes of the LAA clusters and the log-transformed cumulative counts of the clusters larger than the given volume are plotted. This plot is well characterized by a linear regression. The fractal dimension D was obtained as an absolute slope of the regression line. D950 and D’25 indicate the fractal D characterizing LAA clusters based on the fixed − 950 HU and the 25th percentile, respectively. d Distributions of the 15th, 25th, and 35th percentiles of the CT values used in analysis 1 (n = 170) are shown