November 04, 2019: Seminar by Dr. Suvadip Mukherjee, Institut Pasteur, Bioimage Analysis Unit, Paris

Generalized Statistical Object Distance Analysis (GSODA) For Object Based Colocalization In Quantitative Microscopy
Dr. Suvadip Mukherjee, Institut Pasteur, Bioimage Analysis Unit, Paris
The analysis of the spatial distribution of molecules and organelles in bioimaging remains a gold-standard for understanding cellular processes at the molecular level. Recent developments in molecule labeling, optics (e.g. full-field OCT for digital pathology, single molecule localization microscopy in fluorescence microscopy, etc.) and automated image analysis of biological objects such as molecular complexes, organelles or cells
etc. call for the development of advanced statistical tools to analyze the objects’ spatial distribution. In that context, we have recently introduced SODA [1] to study spatial relations between two (or more) spatial point processes. However, in many applications as illustrated, the reduction of analyzed objects  to the position of their center-of-mass effects the robustness of analysis. To generalize SODA (G-SODA) and extend its applicability to more generic region based colocalization studies, we leverage the flexibility of level sets [2] to implicitly embed closed regions, and propose a mathematical model to analytically estimate the model parameters to describe the colocalization properties of the underlying random process. GSODA preserves the statistical characteristics of SODA, but extends its use to a wider gamut of colocalization studies. Furthermore, by restricting the level set function to the region of interest by using suitable boundary conditions, GSODA eliminates the need to explicitly correct for edge-artifacts, which could be challenging to compute for complex shapes. Finally, due to the recent advances in real-time polygon based digital morphology, GSODA allows computationally efficient solution for processing big data in bio-image informatics applications. This makes the proposed solution both robust and computationally efficient, and provides a generic approachto colocalization analysis for a variety of bioimaging studies in fluorescent microscopy, super resolution imaging, and histopathology.
Fig. 1 : Two potential applications of object based colocalization is presented here. The first example (a1) involves estimating protein colocalization in super-resolution microscopy. In (a2), an example is taken from histopathology where the objective is to study the spatial colocalization of the immune cells (points) to the tumor regions (a3).
[1] T. Lagache et al., "Mapping molecular assemblies with fluorescence microscopy and object-based spatial statistics," Nature communications, vol. 9.1, p. 698, 2018.
[2] J. A. Sethian, “Level set methods and fast marching methods: evolving interfaces in computational geometry, fluid mechanics, computer vision, and materials science”, Cambridge University Press, 1999.


All are cordially invited.

Dipti Prasad Mukherjee
Head, ECSU