Skip to main content
. 2018 Jan 12;8:646. doi: 10.1038/s41598-017-19101-3

Figure 2.

Figure 2

A visual representation of the inter-observer and method-observer segmentation performance for solid core, vessel, and ground glass segmentation. From top to bottom, the figures show the results in terms of sensitivity, precision, and Dice similarity score, respectively. The results are presented as the mean (indicated by the colored diamond) and the standard deviation (indicated by the whiskers). The left side of each figure shows the mean and standard deviation derived from nodules for which there was either agreement or disagreement on the presence of a class. The right side of each figure shows the mean and standard deviation derived from only the nodules for which there was agreement on the presence of a class. The number of nodules from which the means and standard deviations were calculated are reported on the top of the whiskers. It should be noted that the experiment that compares three dimensional annotations of observer 3 to the results of the method, i.e. method vs O3 (3D), was performed on a subset of 16 nodules.