Fig. 3

K-means and hierarchical cluster analysis of liver segments. Liver segments were grouped by k-means clustering into four clusters. The optimal number of clusters was determined with the silhouette method. The heatmap was constructed by hierarchical clustering of liver segments (rows) and texture parameters (columns) in each of the four k-means clusters. For drawing the dendrograms, the Pearson correlation was used as a distance metric, and linkage was determined by the Ward. D method. The distribution of clinical and technical variables was also compared among the clusters (color bars)