Skip to main content

Table 1 Classification performance of the individual classifier for each lung region using the full feature set

From: The development and evaluation of a computerized diagnosis scheme for pneumoconiosis on digital chest radiographs

Lung region Training dataset Test dataset
  AUC Spe Sen Acc AUC Spe Sen Acc
(a) 0.932 0.885 0.852 0.877 0.855 0.805 0.776 0.791
(b) 0.961 0.991 0.682 0.891 0.884 0.928 0.699 0.826
(c) 0.996 1.000 0.976 0.990 0.688 0.788 0.474 0.682
(d) 0.824 0.833 0.655 0.778 0.742 0.763 0.594 0.710
(e) 0.924 0.902 0.766 0.854 0.834 0.838 0.699 0.791
(f) 0.722 0.570 0.791 0.648 0.563 0.531 0.728 0.592
Mean 0.893 0.864 0.787 0.840 0.761 0.776 0.662 0.732
SD 0.093 0.144 0.107 0.106 0.111 0.121 0.100 0.080
  1. AUC = area under ROC curve; Spe = specificity; Sen = sensitivity; Acc = accuracy; SD = standard deviation.
  2. Lung regions (a-c) correspond to upper, middle and lower lung field of the right lung, and region (d-f) correspond to upper, middle and lower lung field of the left lung, respectively.