During the training of the INeN, the effect of the threshold value on the segmentation results obtained by the first feature extraction method is examined with the values 4500 and 3000. Five and
seven labels are obtained for 4500 and 3000 values, respectively. Since the ground-truth information
does not exist, each label is used to represent a different candidate class. Fig. 5(a) and (b) show the
labels of the nodes of the INeN for 4500 and 3000 values, respectively. Two threshold values
are selected as 20000 and 18000 for the second feature extraction method. Five and seven classes are obtained for 20000 and 18000 values, respectively. Fig. 5(c) and (d) show the labels of the nodes of the INeN for 20000 and 18000 values, respectively.