We visualize the predictions of multiple machine learning models to help biologists as they interactively make decisions about cell lineage-the development of a (plant) embryo from a single ovum cell. Based on a confocal microscopy dataset, traditionally biologists manually constructed the cell lineage, starting from this observation and reasoning backward in time to establish their inheritance. To speed up this tedious process, we make use of machine learning (ML) models trained on a database of manually established cell lineages to assist the biologist in cell assignment. Most biologists, however, are not familiar with ML, nor is it clear to them which model best predicts the embryo's development. We thus have developed a visualization system that is designed to support biologists in exploring and comparing ML models, checking the model predictions, detecting possible ML model mistakes, and deciding on the most likely embryo development. To evaluate our proposed system, we deployed our interface with six biologists in an observational study. Our results show that the visual representations of machine learning are easily understandable, and our tool, LineageD+, could potentially increase biologists' working efficiency and enhance the understanding of embryos.

Abstract We describe LineageD—a hybrid web‐based system to predict, visualize, and interactively adjust plant embryo cell lineages. Currently, plant biologists explore the development of an embryo and its hierarchical cell lineage manually, based on a 3D dataset that represents the embryo status at one point in time. This human decision‐making process, however, is time‐consuming, tedious, and error‐prone due to the lack of integrated graphical support for specifying the cell lineage. To fill this gap, we developed a new system to support the biologists in their tasks using an interactive combination of 3D visualization, abstract data visualization, and correctable machine learning to modify the proposed cell lineage. We use existing manually established cell lineages to obtain a neural network model. We then allow biologists to use this model to repeatedly predict assignments of a single cell division stage. After each hierarchy level prediction, we allow them to interactively adjust the machine learning based assignment, which we then integrate into the pool of verified assignments for further predictions. In addition to building the hierarchy this way in a bottom‐up fashion, we also offer users to divide the whole embryo and create the hierarchy tree in a top‐down fashion for a few steps, improving the ML‐based assignments by reducing the potential for wrong predictions. We visualize the continuously updated embryo and its hierarchical development using both 3D spatial and abstract tree representations, together with information about the model's confidence and spatial properties. We conducted case study validations with five expert biologists to explore the utility of our approach and to assess the potential for LineageD to be used in their daily workflow. We found that the visualizations of both 3D representations and abstract representations help with decision making and the hierarchy tree top‐down building approach can reduce assignments errors in real practice.

We report on a controlled user study in which we investigated and compared three selection techniques in discovering and traversing 3D objects in densely packed environments. We apply this to cell division history marking as required by plant biologists who study the development of embryos, for whom existing selection techniques do not work due to the occlusion and tight packing of the cells to be selected. We specifically compared a list-based technique with an additional 3D view, a 3D selection technique that relies on an exploded view, and a combination of both techniques. Our results indicate that the combination was most preferred. List selection has advantages for traversing cells, while we did not find differences for surface cells. Our participants appreciated the combination because it supports discovering 3D objects with the 3D explosion technique while using the lists to traverse 3D cells.

Plant tissue architecture and organ morphogenesis rely on the proper orientation of cell divisions. Previous attempts to predict division planes from cell geometry in plants mostly focused on 2D symmetric divisions. Using the stereotyped division patterns of Arabidopsis thaliana early embryogenesis, we investigated geometrical principles underlying plane selection in symmetric and in asymmetric divisions within complex 3D cell shapes. Introducing a 3D computational model of cell division, we show that area minimization constrained on passing through the cell centroid predicts observed divisions. Our results suggest that the positioning of division planes ensues from cell geometry and gives rise to spatially organized cell types with stereotyped shapes, thus underlining the role of self-organization in the developing architecture of the embryo. Our data further suggested the rule could be interpreted as surface minimization constrained by the nucleus position, which was validated using live imaging of cell divisions in the stomatal cell lineage.

UNLABELLED: We developed a system for managing data from two-dimensional electrophoresis-based proteomic experiments. Named PARIS, the system stores gel image and information about experiments and analysis procedures, allows the user to search and navigate in genomic and proteomic data, supports visual verification and validation of the analysis results, and provides tools for cross multi-experiment and multi-experimenter data validation and exploration. AVAILABILITY: The software is freely available from http://www.inra.fr/bia/J/imaste/Projets/PARIS/index.html

One essential assumption used in object detection and labeling by imaging is that the photometric properties of the object are homogeneous. This homogeneousness requirement is often violated in microscopy imaging. Classical methods are usually of high computational cost and fail to give a stable solution. This paper presents a low computational complexity and robust method for three-dimensional (3-D) biological object detection and labeling. The developed approach is based on a statistical, nonparametric framework. Image is first divided into regular nonoverlapped regions and each region is evaluated according to a general photometric variability model. The regions not consistent with this model are considered as aberration in the data and excluded from the analysis procedure. Simultaneously, the interior parts of the object are detected, they correspond to regions where the supposed model is valid. In the second stage, the valid regions from a same object are merged together depending on a set of hypotheses. These hypotheses are generated by taking into account photometric and geometric properties of objects of interest and the merging is achieved according to an iterative algorithm. The approach has been applied in investigations of spatial distribution of nuclei within colonic glands of rats observed with the help of confocal fluorescence microscopy.

The paper addresses the problem of both prior modeling and object labeling in multivariate microscopy imaging. We make use of a statistical, nonparametric framework to formulate the prior knowledge on microscopy imaging and a model validation technique to achieve the object detection and labeling goal. The approach has been applied in investigations of spatial distribution of nuclei within the colonic glands of rats observed with the help of confocal fluorescence microscopy.

An energy model based approach for estimating object boundaries is presented. We study a particular energy whose minimizer can be determined. The method estimates the unknown number of objects and draws object boundaries by selecting the "best" level lines computed from level sets of the original image. Unlike previous standard methods, the proposed method does not require iteration for minimizing the energy. In addition, our segmentation algorithm combines anisotropic diffusion based regularization with level line selection to extract smooth object boundaries. Experimental results on 2D biomedical and meteorological images are reported.

One-dimensional electrophoresis is widely used for the estimation of molecular weight, genotype determination, identification of varieties and DNA sequencing. Many tools have been developed in numerous laboratories in order to perform semi-automatic and sometimes automatic analysis of gels. A one-dimensional electrophoregram consists of lanes and bands; pertinent information is essentially one-dimensional and, for each lane, entirely contained in an optical density profile. Detection of lanes and bands, quantitative analysis and fragment size or molecular weight calculation are provided by some software packages. Here we present our approach to one-dimensional gel analysis using digitized images. We concentrate on the correction of global distortion that may result from inhomogeneous electric field through the gel and on the interpretation of gels in the specific case of genetic studies of families. The package contains most of the usual facilities found in other special-purpose software for one-dimensional gel analysis but particular attention is given here to distortions and interpretation.