Fig. 6.
Clustering of biopsies and principal component analysis of histological readouts. (a) K-means clustering of histological readouts identified two clusters. Cluster 2 contains all MC0 biopsies, 4 of 6 MC2 biopsies and 3 of 8 MC1 biopsies. Cluster 1 only contained MC1 or MC2 biopsies. (b) Principal component analysis of histological readouts of biopsies from control bone marrow (MC0), Modic type 1 change (MC1) and Modic type 2 change (MC2) bone marrow. Principal component 1 (PC1) and principal component 2 (PC2) explained over 65% of the variance. Loadings of the histological readouts are presented as arrows. Violin plots on top and the right show distribution of the biopsies on PC1 and PC2.