Segmentation of Dynamic Total-Body [¹⁸F]-FDG PET Images Using Unsupervised Clustering

Figure 2

Mean Jaccard index over 10 images (-axis) as a function of the number of clusters (-axis) for (a) -means with principal component analysis (PCA), (b) -means with independent component analysis (ICA), and (c) Gaussian mixture models (GMM). The accuracy represented by the mean Jaccard index is calculated separately for heart (red lines), brain (blue lines), and kidney (cyan lines). The selected number of clusters for further analyses is indicated with a black vertical line.

(a)

(b)

(c)