• Home
• Partners
  • University of Bremen
  • EPFL
  • CNRS-UP
  • Tel Aviv University
  • University of Applied Sciences Aschaffenburg
  • University of Vienna
  • SagivTech Ltd.
  • Steinbeis Innovation gGmbH
  • European Research Services GmbH
  • GENESIS S.A.
• Objectives
  • Localization Measures
  • Construction of Minimizers
  • Frames and Signal Expansions
  • Baseline Algorithms
  • GPU Implementation
  • Audio Signal Analysis, Sound Design
  • 3D MALDI Imaging
  • Dissemination
• Publications
  • 2013
  • 2012
  • 2011
  • 2010
• UNLocBoX
• Meetings and Events
• Internal
• Contact

3D MALDI Imaging

The development of new drugs (pharma industry) as well as the search for determining biomarkers for specific diseases (medical diagnosis) rely on a detailed knowledge of the metabolism/proteom of the organism under investigation. Tomographic methodologies (CT, MR, PET, SPECT, PET-MR) rely on specific tracer substances for retrieving at best a partial information of the metabolism.

In contrast, the full metabolic information is contained in the mass-spectrum obtained by mass spectroscopy (MS) experiments. However, the classical MS spectrum obtained from a single tissue sample allows no spatial discrimination. Our proposed proof of concept goes beyond these limits and aims at developing a 3D MALDI MS technology. If successful, our newly developed algorithms will turn the classical 1D-mass spectroscopy (MS) into a 3D-imaging modality.

The recently developed 2D MALDI (matrix assisted laser deisotoping)-MS imaging allows to collect up to more than 10⁴ spectra from a single tissue sample. Our approach is extended: We use the 2D-MALDI data sample for several slices of an object.

We have the following aims:

• Data preprocessing: 3D MALDI data (approx. 50 slices, 100x200 data points per slice, one MS-spectrum with 10⁴ spectral values per data point; approx. 10¹⁰ data values in total) has different noise characteristics and resolutions along the different coordinate axis (high resolution of MS spectra, reduced resolution of sampling points on each tissue slice, poor resolution between slices), we aim at exploiting advanced adaptive denoising techniques for an efficient preprocessing.
• Reduction of MALDI data: We aim at applying optimally localized function systems (scale-space transforms) followed by a “peak picking” routine (selecting characteristic values).
• Classification: We will apply fast hierarchical clustering routines as well as ultra-efficient modification of the k-means algorithm. This results in a 3D-spatially resolved grouping of similar points in different classes. Similarity is measured by similar information content in the related MS-spectra. Hence, different classes represent points with different metabolic behaviour (class imaging).
• Visualization: We will develop visualization routines (volume rendering, partial registration) for displaying the classification result in order to present the metabolic information in a ready to use format for a pharmaceutical/medical expert. We will adapt the software standards (ClinProTools, flexImaging) used by Bruker Daltonik for displaying MALDI information.