Skip to main content
. 2020 Nov 9;10:19388. doi: 10.1038/s41598-020-76389-4

Figure 4.

Figure 4

Schematic illustration of L2-constrained softmax and LOF. For each case, a feature descriptor is obtained from the DCNN. The left part shows a plot of the feature descriptors. LOF can be used to determine the threshold of benign tumors (inside the circle) based on the local density of the training dataset, and discriminate whether the test dataset are normal (benign) or abnormal (malignant). LOF local outlier factor.