Seminarium Zakładu Biofizyki

Due to their size and complexity, Mass Spectrometry Imaging (MSI) data sets often require computational approaches for statistical analyses.One of the typical steps in such analyses is image segmentation, which attempts to highlight regions with characteristic chemical compositions.To improve the quality of MSI data segmentation, we have combined the capabilities of two recently developed algorithms: masserstein and Spatial-DGMM. Separating the overlapping envelopes with masserstein allows us to discover the true regions of analyte concentrations, while incorporating the spatial dependencies between pixels with spatial-DGMM allows us to obtain segments that are more spatially consistent, and therefore easier to interpret and more meaningful biologically.