Figure 14.

Example of fuzzy clustering. Original caption: Identification of intratumor heterogeneity in imaging MS datasets by unsupervised multivariate analysis. Target images were created that contain the unrefined heterogeneity in an imaging MS dataset of intermediate grade myxofibrosarcoma. The outputs of principal component analysis, non‐negative matrix factorization, maximum autocorrelation factor analysis, fuzzy $c$‐means, and probabilistic latent semantic analysis were then examined to identify the components that contained the heterogeneity of the target images. The digit contained in the upper right corner of the component mass spectra indicates which component was used. Most data analysis techniques could reproduce the target images. When the component images reproduced the target images it can be seen that the component mass spectra contain the same peptide and protein ions. Note: PCA and MAF can have negative values, consequently the background surrounding the tissue (defined as zero intensity) can change color. y axis labels, a.u., arbitrary units. Source: Adapted from Jones et al., 2011, Figure 7, under CC‐BY License. MS, mass spectrometry. [Color figure can be viewed at wileyonlinelibrary.com]