Figure 5
From: Statistical physics approach to quantifying differences in myelinated nerve fibers
Unsupervised class recognition.
Dendrogram of an agglomerative hierarchical clustering algorithm49 applied to all the fornix samples belonging to the 6 different subjects, described only by their fraction of occupied area and effective local density. Samples belonging to the young subjects are written in blue while samples of the old subjects are written in red. This algorithm is a systematic procedure to characterize the scatter plot seen in Fig. 4, without knowing the class of the samples, by grouping samples with similar characteristics together. This grouping (or clustering) is done by minimizing the variance of the cluster being merged, with that minimized variance shown as the distance of that merged group.
