Thick specimens, as encountered in cryo-scanning transmission electron tomography, offer special challenges to conventional reconstruction workflows. The visibility of features, including gold nanoparticles introduced as fiducial markers, varies strongly through the tilt series. As a result, tedious manual refinement may be required in order to produce a successful alignment. Information from highly tilted views must often be excluded to the detriment of axial resolution in the reconstruction. We introduce here an approach to tilt series alignment based on identification of fiducial particle clusters that transform coherently in rotation, essentially those that lie at similar depth. Clusters are identified by comparison of tilted views with a single untilted reference, rather than with adjacent tilts. The software, called ClusterAlign, proves robust to poor signal to noise ratio and varying visibility of the individual fiducials and is successful in carrying the alignment to the ends of the tilt series where other methods tend to fail. ClusterAlign may be used to generate a list of tracked fiducials, to align a tilt series, or to perform a complete 3D reconstruction. Tools to evaluate alignment error by projection matching are included. Execution involves no manual intervention, and adherence to standard file formats facilitates an interface with other software, particularly IMOD/etomo, tomo3d, and tomoalign.
This paper presents a deep-learning-based workflow to detect synapses and predict their neurotransmitter type in the primitive chordate Ciona intestinalis (Ciona) electron microscopic (EM) images. Identifying synapses from EM images to build a full map of connections between neurons is a labor-intensive process and requires significant domain expertise. Automation of synapse classification would hasten the generation and analysis of connectomes. Furthermore, inferences concerning neuron type and function from synapse features are in many cases difficult to make. Finding the connection between synapse structure and function is an important step in fully understanding a connectome. Class Activation Maps derived from the convolutional neural network provide insights on important features of synapses based on cell type and function. The main contribution of this work is in the differentiation of synapses by neurotransmitter type through the structural information in their EM images. This enables the prediction of neurotransmitter types for neurons in Ciona, which were previously unknown. The prediction model with code is available on GitHub.
Fluorescence microscopy techniques have experienced a substantial increase in the visualization and analysis of many biological processes in life science. We describe a semiautomated and versatile tool called Cell-TypeAnalyzer to avoid the time-consuming and biased manual classification of cells according to cell types. It consists of an open-source plugin for Fiji or ImageJ to detect and classify cells in 2D images. Our workflow consists of (a) image preprocessing actions, data spatial calibration, and region of interest for analysis; (b) segmentation to isolate cells from background (optionally including user-defined preprocessing steps helping the identification of cells); (c) extraction of features from each cell; (d) filters to select relevant cells; (e) definition of specific criteria to be included in the different cell types; (f) cell classification; and (g) flexible analysis of the results. Our software provides a modular and flexible strategy to perform cell classification through a wizard-like graphical user interface in which the user is intuitively guided through each step of the analysis. This procedure may be applied in batch mode to multiple microscopy files. Once the analysis is set up, it can be automatically and efficiently performed on many images. The plugin does not require any programming skill and can analyze cells in many different acquisition setups.
Advances in tissue engineering for cardiac regenerative medicine require cellular-level understanding of the mechanism of cardiac muscle growth during embryonic developmental stage. Computational methods to automatize cell segmentation in 3D and deliver accurate, quantitative morphology of cardiomyocytes, are imperative to provide insight into cell behavior underlying cardiac tissue growth. Detecting individual cells from volumetric images of dense tissue, poised with low signal-to-noise ratio and severe intensity in homogeneity, is a challenging task. In this article, we develop a robust segmentation tool capable of extracting cellular morphological parameters from 3D multifluorescence images of murine heart, captured via light-sheet microscopy. The proposed pipeline incorporates a neural network for 2D detection of nuclei and cell membranes. A graph-based global association employs the 2D nuclei detections to reconstruct 3D nuclei. A novel optimization embedding the network flow algorithm in an alternating direction method of multipliers is proposed to solve the global object association problem. The associated 3D nuclei serve as the initialization of an active mesh model to obtain the 3D segmentation of individual myocardial cells. The efficiency of our method over the state-of-the-art methods is observed via various qualitative and quantitative evaluation.
Detection of RNA spots in single-molecule fluorescence in-situ hybridization microscopy images remains a difficult task, especially when applied to large volumes of data. The variable intensity of RNA spots combined with the high noise level of the images often requires manual adjustment of the spot detection thresholds for each image. In this work, we introduce DeepSpot, a Deep Learning-based tool specifically designed for RNA spot enhancement that enables spot detection without the need to resort to image per image parameter tuning. We show how our method can enable downstream accurate spot detection. DeepSpot's architecture is inspired by small object detection approaches. It incorporates dilated convolutions into a module specifically designed for context aggregation for small object and uses Residual Convolutions to propagate this information along the network. This enables DeepSpot to enhance all RNA spots to the same intensity, and thus circumvents the need for parameter tuning. We evaluated how easily spots can be detected in images enhanced with our method by testing DeepSpot on 20 simulated and 3 experimental datasets, and showed that accuracy of more than 97% is achieved. Moreover, comparison with alternative deep learning approaches for mRNA spot detection (deepBlink) indicated that DeepSpot provides more precise mRNA detection. In addition, we generated single-molecule fluorescence in-situ hybridization images of mouse fibroblasts in a wound healing assay to evaluate whether DeepSpot enhancement can enable seamless mRNA spot detection and thus streamline studies of localized mRNA expression in cells.
To overcome the physical barriers caused by light diffraction, super-resolution techniques are often applied in fluorescence microscopy. State-of-the-art approaches require specific and often demanding acquisition conditions to achieve adequate levels of both spatial and temporal resolution. Analyzing the stochastic fluctuations of the fluorescent molecules provides a solution to the aforementioned limitations, as sufficiently high spatio-temporal resolution for live-cell imaging can be achieved using common microscopes and conventional fluorescent dyes. Based on this idea, we present COL0RME, a method for covariance-based super-resolution microscopy with intensity estimation, which achieves good spatio-temporal resolution by solving a sparse optimization problem in the covariance domain and discuss automatic parameter selection strategies. The method is composed of two steps: the former where both the emitters' independence and the sparse distribution of the fluorescent molecules are exploited to provide an accurate localization; the latter where real intensity values are estimated given the computed support. The paper is furnished with several numerical results both on synthetic and real fluorescence microscopy images and several comparisons with state-of-the art approaches are provided. Our results show that COL0RME outperforms competing methods exploiting analogously temporal fluctuations; in particular, it achieves better localization, reduces background artifacts, and avoids fine parameter tuning.
Eukaryotic cells are constantly subject to DNA damage, often with detrimental consequences for the health of the organism. Cells mitigate this DNA damage through a variety of repair pathways involving a diverse and large number of different proteins. To better understand the cellular response to DNA damage, one needs accurate measurements of the accumulation, retention, and dissipation timescales of these repair proteins. Here, we describe an automated implementation of the "quantitation of fluorescence accumulation after DNA damage" method that greatly enhances the analysis and quantitation of the widely used technique known as laser microirradiation, which is used to study the recruitment of DNA repair proteins to sites of DNA damage. This open-source implementation ("qFADD.py") is available as a stand-alone software package that can be run on laptops or computer clusters. Our implementation includes corrections for nuclear drift, an automated grid search for the model of a best fit, and the ability to model both horizontal striping and speckle experiments. To improve statistical rigor, the grid-search algorithm also includes automated simulation of replicates. As a practical example, we present and discuss the recruitment dynamics of the early responder PARP1 to DNA damage sites.
Fluorescence microscopy is a critical tool for cell biology studies on bacterial cell division and morphogenesis. Because the analysis of fluorescence microscopy images evolved beyond initial qualitative studies, numerous images analysis tools were developed to extract quantitative parameters on cell morphology and organization. To understand cellular processes required for bacterial growth and division, it is particularly important to perform such analysis in the context of cell cycle progression. However, manual assignment of cell cycle stages is laborious and prone to user bias. Although cell elongation can be used as a proxy for cell cycle progression in rod-shaped or ovoid bacteria, that is not the case for cocci, such as Staphylococcus aureus. Here, we describe eHooke, an image analysis framework developed specifically for automated analysis of microscopy images of spherical bacterial cells. eHooke contains a trained artificial neural network to automatically classify the cell cycle phase of individual S. aureus cells. Users can then apply various functions to obtain biologically relevant information on morphological features of individual cells and cellular localization of proteins, in the context of the cell cycle.