Alexander Broersen defended his PhD thesis on biomedical visualization on 3 March.
How can you find a small, tiny detail in huge databases with millions of research data? This 'needle in a haystack' can be visualized by feature extraction - a visualization technique further developed by Alexander Broersen of the Centrum Wiskunde & Informatica (CWI) in Amsterdam. On 3 March 2009 he received his PhD at the Eindhoven University of Technology with his thesis Feature Visualization in Large Scale Imaging Mass Spectrometry Data. The results make it possible to analyse large numbers of biomedical images from mass spectrometry better and faster.
Chicken embryo
Broersen cooperated with researchers from the FOM Institute for Atomic and Molecular Physics (AMOLF), where research is conducted on the molecular composition of tissues, using mass spectrometry. Goal is to see whether sick and healthy cells can be distinguished, for instance with breast cancer or Alzheimer's disease. A few minutes of measuring time already renders thousands of images (spectra) and many gigabytes of information. Until recently, these data were analysed by hand. Broersen developed a technique to automatically detect certain characteristics, and to visualize them. An example is the visualization of a chicken embryo, not only showing lungs, aorta and bone marrow but also the cells inside the bone marrow. A 'regular' X-ray only shows variations in the density of the tissue, but this research shows a distribution of all molecules varying over the cells.
Special about this method is that it is not a specific search after one substance or characteristic (like cholesterol), but after all deviations. This way unknown correlations between substances can de detected. Broersen processes the images into a kind of information cubes that are processed further with principal component analysis. This takes a lot of calculating power and storage space, which can be supplied by computer grids in the future. In the meantime, AMOLF has taken the results in use to analyse actual data.
Work was done in the Virtual Laboratory for e-Science project (www.vl-e.nl). It was supported by a Bsik grant of OC&W and is part of the ICT innovation programme of the Ministry of Economic Affairs. Supervisors: Prof. R. van Liere (CWI and TU/e) and Prof. R.M.A. Heeren (AMOLF/UU).
More information:
Visualization and 3D Interfaces onderzoeksgroep
Image: visualization of a chicken embryo
