Figure 3. Schematic explanation of model simulation.

Representative CT images from current smokers are shown as binary images, in which black and gray represent pixels with low attenuation and normal density, respectively (A). One pixel was selected from all pixels with normal density and changed to new low attenuation according to two models. Probabilities of selecting pixels between extant low attenuation clusters leading to their coalescence were determined in advance for both models. According to coalescence rates, pixels with normal density were changed into new pixels with low attenuation leading to coalescence or appearance/enlargement of low attenuation clusters (yellow and green pixels, respectively). New, low-attenuation pixels were randomly selected from among all pixels with normal density in the random model (B). Normal density pixels were identified (C) among local advanced emphysematous lesions (blue pixels) in the damage-dependent model, and then new low attenuation pixels were randomly selected from these relatively damaged local lesions (D).