Figure 2 | Scientific Reports

Figure 2

From: The degradation of performance of a state-of-the-art skin image classifier when applied to patient-driven internet search

Figure 2

Binary classification for determining malignancy using the hospital (SNU and Edinburgh) datasets. (a) TEST = Edinburgh dataset (1,300 images). (b) TEST = SNU dataset (2,201 images). (c) TEST = SNU public subset dataset (240 images), Red dot (TH1)-The algorithm at the high sensitivity threshold, Blue dot (TH2)-The algorithm at the high-specificity threshold, + − Average of dermatologists, residents, and laypersons in the previous study17. The mean sensitivity/specificity using 240 test images was adapted, and the result of the reader study is available at https://doi.org/10.6084/m9.figshare.6454973, The shaded area indicates the 95% confidence interval. The TeleDerm dataset was excluded in this malignancy analysis because it includes only four malignancies.

Back to article page