Peter Serocka’s research while affiliated with Shanghai Institute of Technical Physics and other places

Unsupervised segmentation of multi-spectral images plays an important role in annotating infrared microscopic images and is an essential step in label-free spectral histopathology. In this context, diverse clustering approaches have been utilized and evaluated in order to achieve segmentations of Fourier Transform Infrared (FT-IR) microscopic images that agree with histopathological characterization. We introduce so-called interactive similarity maps as an alternative annotation strategy for annotating infrared microscopic images. We demonstrate that segmentations obtained from interactive similarity maps lead to similarly accurate segmentations as segmentations obtained from conventionally used hierarchical clustering approaches. In order to perform this comparison on quantitative grounds, we provide a scheme that allows to identify non-horizontal cuts in dendrograms. This yields a validation scheme for hierarchical clustering approaches commonly used in infrared microscopy. We demonstrate that interactive similarity maps may identify more accurate segmentations than hierarchical clustering based approaches, and thus are a viable and due to their interactive nature attractive alternative to hierarchical clustering. Our validation scheme furthermore shows that performance of hierarchical two-means is comparable to the traditionally used Ward's clustering. As the former is much more efficient in time and memory, our results suggest another less resource demanding alternative for annotating large spectral images.

Functional super-resolution (fSR) microscopy is based on the automated toponome imaging system (TIS). fSR-TIS provides insight into the myriad of different cellular functionalities by direct imaging of large subcellular protein networks in morphologically intact cells and tissues, referred to as the toponome. By cyclical fluorescence imaging of at least 100 molecular cell components, fSR-TIS overcomes the spectral limitations of fluorescence microscopy, which is the essential condition for the detection of protein network structures in situ/in vivo. The resulting data sets precisely discriminate between cell types, subcellular structures, cell states and diseases (fSR). With up to 16 bits per protein, the power of combinatorial molecular discrimination (PCMD) is at least 2(100) per subcellular data point. It provides the dimensionality necessary to uncover thousands of distinct protein clusters including their subcellular hierarchies controlling protein network topology and function in the one cell or tissue section. Here we review the technology and findings showing that functional protein networks of the cell surface in different cancers encompass the same hierarchical and spatial coding principle, but express cancer-specific toponome codes within that scheme (referred to as TIS codes). Findings suggest that TIS codes, extracted from large-scale toponome data, have the potential to be next-generation biomarkers because of their cell type and disease specificity. This is functionally substantiated by the observation that blocking toponome-specific lead proteins results in disassembly of molecular networks and loss of function.

One of the problems arising when exploring toponome or other multivariate-image data is the following: Given a family of n gray-value images of, e.g., a given sample of cell tissue, indexed by a collection of n proteins under investigation (so-called MELK data) — each single image representing the varying local concentration of one of those n proteins at the various sites (pixels) of the given sample, how should one quantify, for any two pixels (or clusters of pixels), the (dis)similarity between the corresponding “vectors” of local protein concentrations in question. Some (dis)similarity mappings defined on \mathbbRn\mathbb{R}^n allowing for fast OpenGL texture mapping turned out to be useful in visual inspection of toponome data. Here, we derive two rather general results regarding similarity and dissimilarity mappings and, as a corollary, the fact that the functions that were used for visual inspection of MELK data are, indeed, metrics. We believe that our results are, however, also of more general interest within the ongoing program of elucidating the structure of metrics from a more abstract point of view.

A new challenge to data visualization has arisen from a new laboratory technique that is capable of imaging a large number of biomedical relevant molecule types in a single tissue probe, termed the Toponome. While aiming at deciphering the biochemical interactions of the molecules, and thus their biological functions as well their roles in diseases, no current methods of image analysis are fully suited for this new quality of high-dimensional image data. To overcome this problem we demonstrate a novel framework for interactive real-time visualization, making use of standard graphics acceleration hardware. We show a sample implementation of a threshold-based visualization technique that is connected to the original work of the Toponome authors, improving it by means of fast user interaction.

We have implemented and investigated a spatial extension of the orig-inal ideal storage model by embedding it in a 2D cellular automaton with a diffusion-like coupling between neighboring cells. The resulting ideal storage cellular automaton model (ISCAM) generates many interesting spatio-temporal patterns, in particular spiral waves that grow and "com-pete" with each other. We study this dynamical behavior both mathemat-ically and computationally, and compare it with similar patterns observed in actual chemical processes. Remarkably, it turned out that one can use such CA for modeling all sorts of complex processes, from phase transition in binary mixtures to using them as a metaphor for cancer onset caused by only one short pulse of 'tissue dis-organzation' (changing e.g. for only one single time step the diffusion coefficient) as hypothesized in recent papers questioning the current gene/genome centric view on cancer onset by AO Ping et al.