Multivariate and/or multidimensional image processing in biomedical applications
J. Angulo (Centre de Morphologie Mathématique – Ecole des Mines de Paris, France) and D. Jeulin (Centre de Morphologie Mathématique – Ecole des Mines de Paris, France)
Nowadays many different modalities are available in medical imaging, including computed tomography (CT) scan, functional or dynamic contrast-enhanced magnetic resonance imaging (fMRI) or (DCE-MRI), positron emission tomography (PET). The 2D/3D + time images produces by these advanced devices are useful for cancer diagnosis, radiotherapy or surgery planning, active study of human brain, tumour angiogenesis quantification, etc. In addition, the most recent microscope systems in biomedical laboratories are based on multi/hyper-spectral imaging for brightfield or fluorescence microscopy.
The high throughput exploitation of these multivariate and/or multidimensional images requires advanced image processing methods and algorithms. To take into account jointly the spatial and the temporal/spectral information as well as the way to combine or to reduce the different temporal/spectral dimensions need adapted mathematical models. Moreover, the extension of standard image processing approaches to 4D images leads to inefficient algorithms in terms of computational requirements (memory overload, time of computation, etc.).
In this framework, the aim of this minisymposium is to draw an overview of some recent developments in the field. We focus in particular on techniques which lie in mathematical morphology, multivariate data analysis, statistical classification, graph-based representations and algorithms, spatio-temporal stochastic modelling, etc.