Robust parametric stabilization of moving cells with intensity correction in light microscopy image sequences

Abstract

International audienceAutomatically stabilizing moving living cells in fluorescence microscopy image sequences is required to attain and analyze the actual displacements of subcellular particles. We have designed a stabilization method which can handle within a single parametric framework, the estimation of the global motion and of the temporal intensity variation (e.g., due to photobleaching effect) that we have to compensate for. We have introduced extended parametric motion-intensity constraints and exploited a robust multiresolution estimation scheme insensitive to local independent motions (outliers). We demonstrate the efficiency and the accuracy of our stabilization method on three challenging cellular events: cell development, endosome displacements, protein recruitment

MISTICA: Minimum Spanning Tree-Based Coarse Image Alignment for Microscopy Image Sequences

“…Next, the transformation for each max Z-projections is applied to all the 2D images belonging to the same Z-stack. This post processing closely resembles the working principle of Poorman’s 3D registration (P3D) method [12] and that briefly mentioned in [11], which also apply in-plane transformations as opposed to full 3D transformations.…”

“…Some papers focus on element matching algorithms, for example feature detection-based methods [13, 14], or intensity-based methods [18], while others focus on transformation models, including linear, elastic [15, 22], spline-based models [1, 17]. More robust methods use an intensity-based model for detection and a parametric motion model for transformation [11], or contour-based model for detection and both a linear and a Navier equation model for transformation [16]. …”

A Guided Tour of Selected Image Processing and Analysis Methods for Fluorescence and Electron Microscopy

“…Applications of global regularized method in biological imaging have recently been investigated in [249], [250], [248], [251]- [254]. Because of possible intensity changes (e.g., photobleaching), the data term needs to be adapted [255]. The data term [256] based on the assumption of conservation of intensity and spatial gradient of the image is typically robust to additive illumination changes, which is necessary for several biological applications [257].…”

Motion compensation in two-photon microscopy temporal series

“…SSD, mutual information). When looking for a global motion, simple transformation types (rigid or affine as in [6]) are generally considered. For a better characterization of spatial dynamics, non-linear transformations may also considered, as in [7].…”

“…It can be considered that the linear registration problem is globally solved, and that published linear registration methods are somehow comparable. However, since some changes, due to development, are to be ignored when registering, robust methods have to be preferred [6].…”