Figure 1.

Unsupervised machine learning methods for exploratory data analysis in IMS. An overview of three reviewed method branches, with application to a MALDI FTICR IMS dataset acquired from rat brain (Verbeeck et al., 2017). (Top) Matrix factorization, with nonnegative matrix factorization as a representative example. (Middle) Clustering analysis, with standard k‐means clustering as a representative example. (Bottom) Manifold learning, with t‐SNE as a representative example. IMS, imaging mass spectrometry; FTICR, Fourier transform ion cyclotron resonance; MALDI, matrix‐assisted laser desorption/ionization; t‐SNE, t‐distributed stochastic neighborhood embedding. [Color figure can be viewed at wileyonlinelibrary.com]