Signal and image processing methods for imaging mass spectrometry data

Abstract

Imaging mass spectrometry (IMS) has evolved as an analytical tool for many biomedical applications. This thesis focuses on algorithms for the analysis of IMS data produced by matrix assisted laser desorption/ionization (MALDI) time-of-flight (TOF) mass spectrometer. IMS provides mass spectra acquired at a grid of spatial points that can be represented as hyperspectral data or a so-called datacube. Analysis of this large and complex data requires efficient computational methods for matrix factorization and for spatial segmentation. In this thesis, state of the art processing methods are reviewed, compared and improved versions are proposed. Mathematical models for peak shapes are reviewed and evaluated. A simulation model for MALDI-TOF is studied, expanded and developed into a simulator for 2D or 3D MALDI-TOF-IMS data. The simulation approach paves way to statistical evaluation of algorithms for analysis of IMS data by providing a gold standard dataset. [...]