CMStatistics 2020: Start Registration
View Submission - CMStatistics
Title: Stochastic modelling for desorption in mass spectrometry imaging Authors:  Xavier Loizeau - National Physical Laboratory (United Kingdom) [presenting]
Rory Steven - National Physical Laboratory (United Kingdom)
Martin Metodiev - National Physical Laboratory and Imperial College London (United Kingdom)
Josephine Bunch - National Physical Laboratory and Imperial College London (United Kingdom)
Abstract: Mass spectrometry imaging (MSI) techniques enable spatially resolved detection of thousands of molecular species from a complex sample, such as biological tissue, in a single experiment. Unfortunately, an MSI dataset does not typically provide a quantitative representation of detected species as physicochemical processes underlying MSI (desorption, and ionisation) are not achieved with equal efficiency for all species, and are highly dependent on the molecules present in any given region, limiting any link between observed signal and local chemical concentration, or quantities of matter. To address this issue, the biological sample of interest is modelled here as a point process and a quantitative imaging experiment is defined as any experiment that is a consistent estimator of the intensity field of this point process. This formulation gives additional meaning to many of the data mining techniques used in the MSI community, and insights on the link between MSI and challenges in modern statistic. In this model, relating the desorbed material in an MSI experiment to the intensity field of interest is done through an ill-posed inverse problem, with an unknown operator that must be estimated through calibration.