Abstract

Introduction: Fluorescence microscopy is essential for studying dynamic events and processes in living multicellular systems like tissues, organs, and organisms. High-throughput microscopies such as spinning disk or light sheet open new possibilities with increased viewfield and/or resolution, but the large 2D/3D data requires computational approaches to enable quantitative analysis among experimental conditions. We focus on computational algorithms to detect and measure adjacent cellular membranes and shape properties from such images. This task becomes challenging since the combination of membrane thinness, intrinsic variability of fluorescent signals and membrane shapes, and rapid changes, have precluded automated approaches so far. Material & Methods: 2D/3D parametric active contours model cell membrane shapes: polygons or meshes extracted from rough binary segmentations, are adjusted iteratively towards an optimized shape, balancing image and geometrical-physical contour features. We developed ALPACA (ALgorithm for Piecewise Adjacent Contour Adjustment), defining distance thresholds to detect and optimize shared and non-shared parametric contour sections. We calibrated, applied, and measured ALPACA upon (i) synthetic images, and spinning disk time-lapse stacks of (ii) polyacrylamide droplets, and (iii) parapineal cell membranes in zebrafish embryos with ground-truth segmentations from developmental biologists. Results: ALPACA contours yield smooth and unique shared contour sections, avoiding adjacent cell overlaps and distancing, and measured areas, perimeters, and curvatures close to the ground-truth. Discussion: We achieved geometrically consistent, pairwise cell segmentations in 2D. Current challenges to address are junction point adjustment for three or more cells, and ALPACA integration with 3D active surface models. Acknowledgements: ANID ICN09_015/ICM-P09-015-F, FONDECYT 1221696/1211988/3220832, FONDEQUIP EQM210020.