Skip to main content
. 2018 Nov 28;13(11):e0206081. doi: 10.1371/journal.pone.0206081

Fig 8. Incorrectly classified visual fields from the BD dataset.

Fig 8

Illustration of two cases where the CNN fails to correctly identify early-glaucomatous visual fields, along with the associated SmoothGrad and Piece-wise saliency maps. MD and sLV values for each case are provided, as well as the probability of being early-glaucomatous estimated by the CNN. The corresponding cut-off values are given in parenthesis for each type of scores.