Search tips
Search criteria

Results 1-25 (2114094)

Clipboard (0)

Related Articles

1.  Automated Processing of Imaging Data through Multi-tiered Classification of Biological Structures Illustrated Using Caenorhabditis elegans 
PLoS Computational Biology  2015;11(4):e1004194.
Quantitative imaging has become a vital technique in biological discovery and clinical diagnostics; a plethora of tools have recently been developed to enable new and accelerated forms of biological investigation. Increasingly, the capacity for high-throughput experimentation provided by new imaging modalities, contrast techniques, microscopy tools, microfluidics and computer controlled systems shifts the experimental bottleneck from the level of physical manipulation and raw data collection to automated recognition and data processing. Yet, despite their broad importance, image analysis solutions to address these needs have been narrowly tailored. Here, we present a generalizable formulation for autonomous identification of specific biological structures that is applicable for many problems. The process flow architecture we present here utilizes standard image processing techniques and the multi-tiered application of classification models such as support vector machines (SVM). These low-level functions are readily available in a large array of image processing software packages and programming languages. Our framework is thus both easy to implement at the modular level and provides specific high-level architecture to guide the solution of more complicated image-processing problems. We demonstrate the utility of the classification routine by developing two specific classifiers as a toolset for automation and cell identification in the model organism Caenorhabditis elegans. To serve a common need for automated high-resolution imaging and behavior applications in the C. elegans research community, we contribute a ready-to-use classifier for the identification of the head of the animal under bright field imaging. Furthermore, we extend our framework to address the pervasive problem of cell-specific identification under fluorescent imaging, which is critical for biological investigation in multicellular organisms or tissues. Using these examples as a guide, we envision the broad utility of the framework for diverse problems across different length scales and imaging methods.
Author Summary
New technologies have increased the size and content-richness of biological imaging datasets. As a result, automated image processing is increasingly necessary to extract relevant data in an objective, consistent and time-efficient manner. While image processing tools have been developed for general problems that affect large communities of biologists, the diversity of biological research questions and experimental techniques have left many problems unaddressed. Moreover, there is no clear way in which non-computer scientists can immediately apply a large body of computer vision and image processing techniques to address their specific problems or adapt existing tools to their needs. Here, we address this need by demonstrating an adaptable framework for image processing that is capable of accommodating a large range of biological problems with both high accuracy and computational efficiency. Moreover, we demonstrate the utilization of this framework for disparate problems by solving two specific image processing challenges in the model organism Caenorhabditis elegans. In addition to contributions to the C. elegans community, the solutions developed here provide both useful concepts and adaptable image-processing modules for other biological problems.
PMCID: PMC4409145  PMID: 25910032
2.  A regression model approach to enable cell morphology correction in high-throughput flow cytometry 
Large variations in cell size and shape can undermine traditional gating methods for analyzing flow cytometry data. Correcting for these effects enables analysis of high-throughput data sets, including >5000 yeast samples with diverse cell morphologies.
The regression model approach corrects for the effects of cell morphology on fluorescence, as well as an extremely small and restrictive gate, but without removing any of the cells.In contrast to traditional gating, this approach enables the quantitative analysis of high-throughput flow cytometry experiments, since the regression model can compare between biological samples that show no or little overlap in terms of the morphology of the cells.The analysis of a high-throughput yeast flow cytometry data set consisting of >5000 biological samples identified key proteins that affect the time and intensity of the bifurcation event that happens after the carbon source transition from glucose to fatty acids. Here, some yeast cells undergo major structural changes, while others do not.
Flow cytometry is a widely used technique that enables the measurement of different optical properties of individual cells within large populations of cells in a fast and automated manner. For example, by targeting cell-specific markers with fluorescent probes, flow cytometry is used to identify (and isolate) cell types within complex mixtures of cells. In addition, fluorescence reporters can be used in conjunction with flow cytometry to measure protein, RNA or DNA concentration within single cells of a population.
One of the biggest advantages of this technique is that it provides information of how each cell behaves instead of just measuring the population average. This can be essential when analyzing complex samples that consist of diverse cell types or when measuring cellular responses to stimuli. For example, there is an important difference between a 50% expression increase of all cells in a population after stimulation and a 100% increase in only half of the cells, while the other half remains unresponsive. Another important advantage of flow cytometry is automation, which enables high-throughput studies with thousands of samples and conditions. However, current methods are confounded by populations of cells that are non-uniform in terms of size and granularity. Such variability affects the emitted fluorescence of the cell and adds undesired variability when estimating population fluorescence. This effect also frustrates a sensible comparison between conditions, where not only fluorescence but also cell size and granularity may be affected.
Traditionally, this problem has been addressed by using ‘gates' that restrict the analysis to cells with similar morphological properties (i.e. cell size and cell granularity). Because cells inside the gate are morphologically similar to one another, they will show a smaller variability in their response within the population. Moreover, applying the same gate in all samples assures that observed differences between these samples are not due to differential cell morphologies.
Gating, however, comes with costs. First, since only a subgroup of cells is selected, the final number of cells analyzed can be significantly reduced. This means that in order to have sufficient statistical power, more cells have to be acquired, which, if even possible in the first place, increases the time and cost of the experiment. Second, finding a good gate for all samples and conditions can be challenging if not impossible, especially in cases where cellular morphology changes dramatically between conditions. Finally, gating is a very user-dependent process, where both the size and shape of the gate are determined by the researcher and will affect the outcome, introducing subjectivity in the analysis that complicates reproducibility.
In this paper, we present an alternative method to gating that addresses the issues stated above. The method is based on a regression model containing linear and non-linear terms that estimates and corrects for the effect of cell size and granularity on the observed fluorescence of each cell in a sample. The corrected fluorescence thus becomes ‘free' of the morphological effects.
Because the model uses all cells in the sample, it assures that the corrected fluorescence is an accurate representation of the sample. In addition, the regression model can predict the expected fluorescence of a sample in areas where there are no cells. This makes it possible to compare between samples that have little overlap with good confidence. Furthermore, because the regression model is automated, it is fully reproducible between labs and conditions. Finally, it allows for a rapid analysis of big data sets containing thousands of samples.
To probe the validity of the model, we performed several experiments. We show how the regression model is able to remove the morphological-associated variability as well as an extremely small and restrictive gate, but without the caveat of removing cells. We test the method in different organisms (yeast and human) and applications (protein level detection, separation of mixed subpopulations). We then apply this method to unveil new biological insights in the mechanistic processes involved in transcriptional noise.
Gene transcription is a process subjected to the randomness intrinsic to any molecular event. Although such randomness may seem to be undesirable for the cell, since it prevents consistent behavior, there are situations where some degree of randomness is beneficial (e.g. bet hedging). For this reason, each gene is tuned to exhibit different levels of randomness or noise depending on its functions. For core and essential genes, the cell has developed mechanisms to lower the level of noise, while for genes involved in the response to stress, the variability is greater.
This gene transcription tuning can be determined at many levels, from the architecture of the transcriptional network, to epigenetic regulation. In our study, we analyze the latter using the response of yeast to the presence of fatty acid in the environment. Fatty acid can be used as energy by yeast, but it requires major structural changes and commitments. We have observed that at the population level, there is a bifurcation event whereby some cells undergo these changes and others do not. We have analyzed this bifurcation event in mutants for all the non-essential epigenetic regulators in yeast and identified key proteins that affect the time and intensity of this bifurcation. Even though fatty acid triggers major morphological changes in the cell, the regression model still makes it possible to analyze the over 5000 flow cytometry samples in this data set in an automated manner, whereas a traditional gating approach would be impossible.
Cells exposed to stimuli exhibit a wide range of responses ensuring phenotypic variability across the population. Such single cell behavior is often examined by flow cytometry; however, gating procedures typically employed to select a small subpopulation of cells with similar morphological characteristics make it difficult, even impossible, to quantitatively compare cells across a large variety of experimental conditions because these conditions can lead to profound morphological variations. To overcome these limitations, we developed a regression approach to correct for variability in fluorescence intensity due to differences in cell size and granularity without discarding any of the cells, which gating ipso facto does. This approach enables quantitative studies of cellular heterogeneity and transcriptional noise in high-throughput experiments involving thousands of samples. We used this approach to analyze a library of yeast knockout strains and reveal genes required for the population to establish a bimodal response to oleic acid induction. We identify a group of epigenetic regulators and nucleoporins that, by maintaining an ‘unresponsive population,' may provide the population with the advantage of diversified bet hedging.
PMCID: PMC3202802  PMID: 21952134
flow cytometry; high-throughput experiments; statistical regression model; transcriptional noise
3.  Quantitative Time-Lapse Fluorescence Microscopy in Single Cells 
The cloning of GFP 15 years ago revolutionized cell biology by permitting visualization of a wide range of molecular mechanisms within living cells. Though initially used to make largely qualitative assessments of protein levels and localizations, fluorescence microscopy has since evolved to become highly quantitative and high-throughput. Computational image analysis has catalyzed this evolution, enabling rapid and automated processing of large datasets. Here we review studies that combine time-lapse fluorescence microscopy and automated image analysis to investigate dynamic events at the single-cell level. We highlight examples where single-cell analysis provides unique mechanistic insights into cellular processes that cannot be otherwise resolved in bulk assays. Additionally, we discuss studies where quantitative microscopy facilitates the assembly of detailed 4D lineages in developing organisms. Finally, we describe recent advances in imaging technology, focusing especially on platforms that allow the simultaneous perturbation and quantitative monitoring of biological systems.
PMCID: PMC3137897  PMID: 19575655
cell-to-cell variability; automated image analysis; lineage construction; microfluidic devices; fluorescent proteins
4.  Clustering phenotype populations by genome-wide RNAi and multiparametric imaging 
How to predict gene function from phenotypic cues is a longstanding question in biology.Using quantitative multiparametric imaging, RNAi-mediated cell phenotypes were measured on a genome-wide scale.On the basis of phenotypic ‘neighbourhoods', we identified previously uncharacterized human genes as mediators of the DNA damage response pathway and the maintenance of genomic integrity.The phenotypic map is provided as an online resource at for discovering further functional relationships for a broad spectrum of biological module
Genetic screens for phenotypic similarity have made key contributions for associating genes with biological processes. Aggregating genes by similarity of their loss-of-function phenotype has provided insights into signalling pathways that have a conserved function from Drosophila to human (Nusslein-Volhard and Wieschaus, 1980; Bier, 2005). Complex visual phenotypes, such as defects in pattern formation during development, greatly facilitated the classification of genes into pathways, and phenotypic similarities in many cases predicted molecular relationships. With RNA interference (RNAi), highly parallel phenotyping of loss-of-function effects in cultured cells has become feasible in many organisms whose genome have been sequenced (Boutros and Ahringer, 2008). One of the current challenges is the computational categorization of visual phenotypes and the prediction of gene function and associated biological processes. With large parts of the genome still being in unchartered territory, deriving functional information from large-scale phenotype analysis promises to uncover novel gene–gene relationships and to generate functional maps to explore cellular processes.
In this study, we developed an automated approach using RNAi-mediated cell phenotypes, multiparametric imaging and computational modelling to obtain functional information on previously uncharacterized genes. To generate broad, computer-readable phenotypic signatures, we measured the effect of RNAi-mediated knockdowns on changes of cell morphology in human cells on a genome-wide scale. First, the several million cells were stained for nuclear and cytoskeletal markers and then imaged using automated microscopy. On the basis of fluorescent markers, we established an automated image analysis to classify individual cells (Figure 1A). After cell segmentation for determining nuclei and cell boundaries (Figure 1C), we computed 51 cell descriptors that quantified intensities, shape characteristics and texture (Figure 1F). Individual cells were categorized into 1 of 10 classes, which included cells showing protrusion/elongation, cells in metaphase, large cells, condensed cells, cells with lamellipodia and cellular debris (Figure 1D and E). Each siRNA knockdown was summarized by a phenotypic profile and differences between RNAi knockdowns were quantified by the similarity between phenotypic profiles. We termed the vector of scores a phenoprint (Figure 3C) and defined the phenotypic distance between a pair of perturbations as the distance between their corresponding phenoprints.
To visualize the distribution of all phenoprints, we plotted them in a genome-wide map as a two-dimensional representation of the phenotypic similarity relationships (Figure 3A). The complete data set and an interactive version of the phenotypic map are available at The map identified phenotypic ‘neighbourhoods', which are characterized by cells with lamellipodia (WNK3, ANXA4), cells with prominent actin fibres (ODF2, SOD3), abundance of large cells (CA14), many elongated cells (SH2B2, ELMO2), decrease in cell number (TPX2, COPB1, COPA), increase in number of cells in metaphase (BLR1, CIB2) and combinations of phenotypes such as presence of large cells with protrusions and bright nuclei (PTPRZ1, RRM1; Figure 3B).
To test whether phenotypic similarity might serve as a predictor of gene function, we focused our further analysis on two clusters that contained genes associated with the DNA damage response (DDR) and genomic integrity (Figure 3A and C). The first phenotypic cluster included proteins with kinetochore-associated functions such as NUF2 (Figure 3B) and SGOL1. It also contained the centrosomal protein CEP164 that has been described as an important mediator of the DNA damage-activated signalling cascade (Sivasubramaniam et al, 2008) and the largely uncharacterized genes DONSON and SON. A second phenotypically distinct cluster included previously described components of the DDR pathway such as RRM1 (Figure 3A–C), CLSPN, PRIM2 and SETD8. Furthermore, this cluster contained the poorly characterized genes CADM1 and CD3EAP.
Cells activate a signalling cascade in response to DNA damage induced by exogenous and endogenous factors. Central are the kinases ATM and ATR as they serve as sensors of DNA damage and activators of further downstream kinases (Harper and Elledge, 2007; Cimprich and Cortez, 2008). To investigate whether DONSON, SON, CADM1 and CD3EAP, which were found in phenotypic ‘neighbourhoods' to known DDR components, have a role in the DNA damage signalling pathway, we tested the effect of their depletion on the DDR on γ irradiation. As indicated by reduced CHEK1 phosphorylation, siRNA knock down of DONSON, SON, CD3EAP or CADM1 resulted in impaired DDR signalling on γ irradiation. Furthermore, knock down of DONSON or SON reduced phosphorylation of downstream effectors such as NBS1, CHEK1 and the histone variant H2AX on UVC irradiation. DONSON depletion also impaired recruitment of RPA2 onto chromatin and SON knockdown reduced RPA2 phosphorylation indicating that DONSON and SON presumably act downstream of the activation of ATM. In agreement to their phenotypic profile, these results suggest that DONSON, SON, CADM1 and CD3EAP are important mediators of the DDR. Further experiments demonstrated that they are also required for the maintenance of genomic integrity.
In summary, we show that genes with similar phenotypic profiles tend to share similar functions. The power of our computational and experimental approach is demonstrated by the identification of novel signalling regulators whose phenotypic profiles were found in proximity to known biological modules. Therefore, we believe that such phenotypic maps can serve as a resource for functional discovery and characterization of unknown genes. Furthermore, such approaches are also applicable for other perturbation reagents, such as small molecules in drug discovery and development. One could also envision combined maps that contain both siRNAs and small molecules to predict target–small molecule relationships and potential side effects.
Genetic screens for phenotypic similarity have made key contributions to associating genes with biological processes. With RNA interference (RNAi), highly parallel phenotyping of loss-of-function effects in cells has become feasible. One of the current challenges however is the computational categorization of visual phenotypes and the prediction of biological function and processes. In this study, we describe a combined computational and experimental approach to discover novel gene functions and explore functional relationships. We performed a genome-wide RNAi screen in human cells and used quantitative descriptors derived from high-throughput imaging to generate multiparametric phenotypic profiles. We show that profiles predicted functions of genes by phenotypic similarity. Specifically, we examined several candidates including the largely uncharacterized gene DONSON, which shared phenotype similarity with known factors of DNA damage response (DDR) and genomic integrity. Experimental evidence supports that DONSON is a novel centrosomal protein required for DDR signalling and genomic integrity. Multiparametric phenotyping by automated imaging and computational annotation is a powerful method for functional discovery and mapping the landscape of phenotypic responses to cellular perturbations.
PMCID: PMC2913390  PMID: 20531400
DNA damage response signalling; massively parallel phenotyping; phenotype networks; RNAi screening
5.  Fast automatic quantitative cell replication with fluorescent live cell imaging 
BMC Bioinformatics  2012;13:21.
live cell imaging is a useful tool to monitor cellular activities in living systems. It is often necessary in cancer research or experimental research to quantify the dividing capabilities of cells or the cell proliferation level when investigating manipulations of the cells or their environment. Manual quantification of fluorescence microscopic image is difficult because human is neither sensitive to fine differences in color intensity nor effective to count and average fluorescence level among cells. However, auto-quantification is not a straightforward problem to solve. As the sampling location of the microscopy changes, the amount of cells in individual microscopic images varies, which makes simple measurement methods such as the sum of stain intensity values or the total number of positive stain within each image inapplicable. Thus, automated quantification with robust cell segmentation techniques is required.
An automated quantification system with robust cell segmentation technique are presented. The experimental results in application to monitor cellular replication activities show that the quantitative score is promising to represent the cell replication level, and scores for images from different cell replication groups are demonstrated to be statistically significantly different using ANOVA, LSD and Tukey HSD tests (p-value < 0.01). In addition, the technique is fast and takes less than 0.5 second for high resolution microscopic images (with image dimension 2560 × 1920).
A robust automated quantification method of live cell imaging is built to measure the cell replication level, providing a robust quantitative analysis system in fluorescent live cell imaging. In addition, the presented unsupervised entropy based cell segmentation for live cell images is demonstrated to be also applicable for nuclear segmentation of IHC tissue images.
PMCID: PMC3359210  PMID: 22292799
6.  Automated High-Content Live Animal Drug Screening Using C. elegans Expressing the Aggregation Prone Serpin α1-antitrypsin Z 
PLoS ONE  2010;5(11):e15460.
The development of preclinical models amenable to live animal bioactive compound screening is an attractive approach to discovering effective pharmacological therapies for disorders caused by misfolded and aggregation-prone proteins. In general, however, live animal drug screening is labor and resource intensive, and has been hampered by the lack of robust assay designs and high throughput work-flows. Based on their small size, tissue transparency and ease of cultivation, the use of C. elegans should obviate many of the technical impediments associated with live animal drug screening. Moreover, their genetic tractability and accomplished record for providing insights into the molecular and cellular basis of human disease, should make C. elegans an ideal model system for in vivo drug discovery campaigns. The goal of this study was to determine whether C. elegans could be adapted to high-throughput and high-content drug screening strategies analogous to those developed for cell-based systems. Using transgenic animals expressing fluorescently-tagged proteins, we first developed a high-quality, high-throughput work-flow utilizing an automated fluorescence microscopy platform with integrated image acquisition and data analysis modules to qualitatively assess different biological processes including, growth, tissue development, cell viability and autophagy. We next adapted this technology to conduct a small molecule screen and identified compounds that altered the intracellular accumulation of the human aggregation prone mutant that causes liver disease in α1-antitrypsin deficiency. This study provides powerful validation for advancement in preclinical drug discovery campaigns by screening live C. elegans modeling α1-antitrypsin deficiency and other complex disease phenotypes on high-content imaging platforms.
PMCID: PMC2980495  PMID: 21103396
7.  An automatic method for robust and fast cell detection in bright field images from high-throughput microscopy 
BMC Bioinformatics  2013;14:297.
In recent years, high-throughput microscopy has emerged as a powerful tool to analyze cellular dynamics in an unprecedentedly high resolved manner. The amount of data that is generated, for example in long-term time-lapse microscopy experiments, requires automated methods for processing and analysis. Available software frameworks are well suited for high-throughput processing of fluorescence images, but they often do not perform well on bright field image data that varies considerably between laboratories, setups, and even single experiments.
In this contribution, we present a fully automated image processing pipeline that is able to robustly segment and analyze cells with ellipsoid morphology from bright field microscopy in a high-throughput, yet time efficient manner. The pipeline comprises two steps: (i) Image acquisition is adjusted to obtain optimal bright field image quality for automatic processing. (ii) A concatenation of fast performing image processing algorithms robustly identifies single cells in each image. We applied the method to a time-lapse movie consisting of ∼315,000 images of differentiating hematopoietic stem cells over 6 days. We evaluated the accuracy of our method by comparing the number of identified cells with manual counts. Our method is able to segment images with varying cell density and different cell types without parameter adjustment and clearly outperforms a standard approach. By computing population doubling times, we were able to identify three growth phases in the stem cell population throughout the whole movie, and validated our result with cell cycle times from single cell tracking.
Our method allows fully automated processing and analysis of high-throughput bright field microscopy data. The robustness of cell detection and fast computation time will support the analysis of high-content screening experiments, on-line analysis of time-lapse experiments as well as development of methods to automatically track single-cell genealogies.
PMCID: PMC3850979  PMID: 24090363
8.  Meaningful Interpretation of Subdiffusive Measurements in Living Cells (Crowded Environment) by Fluorescence Fluctuation Microscopy 
In living cell or its nucleus, the motions of molecules are complicated due to the large crowding and expected heterogeneity of the intracellular environment. Randomness in cellular systems can be either spatial (anomalous) or temporal (heterogeneous). In order to separate both processes, we introduce anomalous random walks on fractals that represented crowded environments. We report the use of numerical simulation and experimental data of single-molecule detection by fluorescence fluctuation microscopy for detecting resolution limits of different mobile fractions in crowded environment of living cells. We simulate the time scale behavior of diffusion times τD(τ) for one component, e.g. the fast mobile fraction, and a second component, e.g. the slow mobile fraction. The less the anomalous exponent α the higher the geometric crowding of the underlying structure of motion that is quantified by the ratio of the Hausdorff dimension and the walk exponent d f /dw and specific for the type of crowding generator used. The simulated diffusion time decreases for smaller values of α ≠ 1 but increases for a larger time scale τ at a given value of α ≠ 1. The effect of translational anomalous motion is substantially greater if α differs much from 1. An α value close to 1 contributes little to the time dependence of subdiffusive motions. Thus, quantitative determination of molecular weights from measured diffusion times and apparent diffusion coefficients, respectively, in temporal auto- and crosscorrelation analyses and from time-dependent fluorescence imaging data are difficult to interpret and biased in crowded environments of living cells and their cellular compartments; anomalous dynamics on different time scales τ must be coupled with the quantitative analysis of how experimental parameters change with predictions from simulated subdiffusive dynamics of molecular motions and mechanistic models. We first demonstrate that the crowding exponent α also determines the resolution of differences in diffusion times between two components in addition to photophyscial parameters well-known for normal motion in dilute solution. The resolution limit between two different kinds of single molecule species is also analyzed under translational anomalous motion with broken ergodicity. We apply our theoretical predictions of diffusion times and lower limits for the time resolution of two components to fluorescence images in human prostate cancer cells transfected with GFP-Ago2 and GFP-Ago1. In order to mimic heterogeneous behavior in crowded environments of living cells, we need to introduce so-called continuous time random walks (CTRW). CTRWs were originally performed on regular lattice. This purely stochastic molecule behavior leads to subdiffusive motion with broken ergodicity in our simulations. For the first time, we are able to quantitatively differentiate between anomalous motion without broken ergodicity and anomalous motion with broken ergodicity in time-dependent fluorescence microscopy data sets of living cells. Since the experimental conditions to measure a selfsame molecule over an extended period of time, at which biology is taken place, in living cells or even in dilute solution are very restrictive, we need to perform the time average over a subpopulation of different single molecules of the same kind. For time averages over subpopulations of single molecules, the temporal auto- and crosscorrelation functions are first found. Knowing the crowding parameter α for the cell type and cellular compartment type, respectively, the heterogeneous parameter γ can be obtained from the measurements in the presence of the interacting reaction partner, e.g. ligand, with the same α value. The product α⋅γ=γ˜ is not a simple fitting parameter in the temporal auto- and two-color crosscorrelation functions because it is related to the proper physical models of anomalous (spatial) and heterogeneous (temporal) randomness in cellular systems. We have already derived an analytical solution for γ˜ in the special case of γ = 3/2 . In the case of two-color crosscorrelation or/and two-color fluorescence imaging (co-localization experiments), the second component is also a two-color species gr, for example a different molecular complex with an additional ligand. Here, we first show that plausible biological mechanisms from FCS/ FCCS and fluorescence imaging in living cells are highly questionable without proper quantitative physical models of subdiffusive motion and temporal randomness. At best, such quantitative FCS/ FCCS and fluorescence imaging data are difficult to interpret under crowding and heterogeneous conditions. It is challenging to translate proper physical models of anomalous (spatial) and heterogeneous (temporal) randomness in living cells and their cellular compartments like the nucleus into biological models of the cell biological process under study testable by single-molecule approaches. Otherwise, quantitative FCS/FCCS and fluorescence imaging measurements in living cells are not well described and cannot be interpreted in a meaningful way.
PMCID: PMC3583073  PMID: 20553227
Anomalous motion; broken ergodicity; Continuous Time Random Walks (CTRW); Continuous Time Random Walks (CTRW) on fractal supports; cellular crowding; Cytoplasmic Assembly of Nuclear RISC; ergodicity; FCS; FCCS; Fluorescence Fluctuation Microscopy; GFP-Ago1; GFP-Ago2; heterogeneity; living cells; meaningful interpretation of subdiffusive measurements; microRNA trafficking; physical model of crowding; physical model of heterogeneity; random walks on fractal supports; resolution limits of measured diffusion times for two components; RNA Activation (RNAa); Single Molecule; Small Activating RNA (saRNA); Temporal autocorrelation; Temporal two-color crosscorrelation; Fluorescence imaging; Time dependence of apparent diffusion coefficients.
9.  rsEGFP2 enables fast RESOLFT nanoscopy of living cells 
eLife  2012;1:e00248.
The super-resolution microscopy called RESOLFT relying on fluorophore switching between longlived states, stands out by its coordinate-targeted sequential sample interrogation using low light levels. While RESOLFT has been shown to discern nanostructures in living cells, the reversibly photoswitchable green fluorescent protein (rsEGFP) employed in these experiments was switched rather slowly and recording lasted tens of minutes. We now report on the generation of rsEGFP2 providing faster switching and the use of this protein to demonstrate 25–250 times faster recordings.
eLife digest
For decades it was assumed that the diffraction of light meant that optical microscopy could not resolve features that were smaller than about the half the wavelength of the light being used to create an image. However, various ‘super-resolution’ methods have allowed researchers to overcome this diffraction limit for fluorescence imaging, which is the most popular form of microscopy used in the life sciences. This approach involves tagging the biomolecules of interest with fluorescent molecules, such as green fluorescent protein (GFP), so that they can be identified in cells. An excitation laser then drives the fluorescent molecule, which is also known as a fluorophore, into an excited state: after a short time, the fluorophore can return to its ground state by releasing a fluorescence photon. Images of the sample are built up by detecting these photons.
In STED super-resolution microscopy a second laser is used to instantly send the molecules from their excited or ‘on’ states back to their ground or ‘off’ states before any fluorescence can occur. The second laser beam is usually shaped like a doughnut, with a small region of low light intensity surrounded by a region of much higher intensity. STED microscopy is able to beat the diffraction limit because the second laser turns all the fluorophores ‘off’ except those in the small sub-wavelength region at the centre of the doughnut. The image is build up by scanning both lasers over the sample so that the small region in which the fluorophores are ‘on’ probes the entire cell.
RESOLFT is a similar technique that employs fluorescent molecules with ‘on’ and ‘off’ times that are much longer than those used in STED microscopy. In particular, RESOLFT uses fluorescent molecules that can be rapidly switched back and forth between long-lived ‘on’ and ‘off’ states many times by the two lasers. The fact that both these states are long-lived states means that RESOLFT requires much lower laser intensities than STED, which makes it attractive for imaging biological samples over large areas or long times.
RESOLFT demonstrated its suitability for bioimaging for the first time last year, with a protein called rsEGFP (reversibly switchable enhanced GFP) being employed as the fluorophore. However, the time needed to switch this protein between the ‘on state’ and the ‘off state’ was relatively long, and it took about an hour to record a typical image. Now, Grotjohann et al. have modified this protein to make a new fluorophore called rsEGFP2 with a shorter switching time, and have used it to image various structures—including Vimentin, a protein that forms part of the cytoskeleton in many cells, and organelles called peroxisomes—inside live mammalian cells. They were able to record these images some 25–250 times faster than would have been possible with previous RESOLFT approaches. The combination of RESOLFT and rsEGFP2 should allow researchers to image a wide variety of structures and processes in living cells that have not been imaged before.
PMCID: PMC3534202  PMID: 23330067
confocal microscopy; fluorescent probes; GFP; nanoscopy; superresolution; live-cell imaging; None
10.  A modular gradient-sensing network for chemotaxis in Escherichia coli revealed by responses to time-varying stimuli 
Combining in vivo FRET with time-varying stimuli, such as steps, ramps, and sinusoids allowed deduction of the molecular mechanisms underlying cellular signal processing.The bacterial chemotaxis pathway can be described as a two-module feedback circuit, the transfer functions of which we have characterized quantitatively by experiment. Model-driven experimental design allowed the use of a single FRET pair for measurements of both transfer functions of the pathway.The adaptation module's transfer function revealed that feedback near steady state is weak, consistent with high sensitivity to shallow gradients, but also strong steady-state fluctuations in pathway output.The measured response to oscillatory stimuli defines the frequency band over which the chemotaxis system can compute time derivatives.
In searching for better environments, bacteria sample their surroundings by random motility, and make temporal comparisons of experienced sensory cues to bias their movement toward favorable directions (Berg and Brown, 1972). Thus, the problem of sensing spatial gradients is reduced to time-derivative computations, carried out by a signaling pathway that is well characterized at the molecular level in Escherichia coli. Here, we study the physiology of this signal processing system in vivo by fluorescence resonance energy transfer (FRET) experiments in which live cells are stimulated by time-varying chemoeffector signals. By measuring FRET between the active response regulator of the pathway CheY-P and its phosphatase CheZ, each labeled with GFP variants, we obtain a readout that is directly proportional to pathway activity (Sourjik et al, 2007). We analyze the measured response functions in terms of mechanistic models of signaling, and discuss functional consequences of the observed quantitative characteristics.
Experiments are guided by a coarse-grained modular model (Tu et al, 2008) of the sensory network (Figure 1), in which we identify two important ‘transfer functions': one corresponding to the receptor–kinase complex, which responds to changes in input ligand concentration on a fast time scale, and another corresponding to the adaptation system, which provides negative feedback, opposing the effect of ligand on a slower time scale. For the receptor module, we calibrate an allosteric MWC-type model of the receptor–kinase complex by FRET measurements of the ‘open-loop' transfer function G([L],m) using step stimuli. This calibration provides a basis for using the same FRET readout (between CheY-P and CheZ) to further study properties of the adaptation module.
It is well known that adaptation in E. coli's chemotaxis system uses integral feedback, which guarantees exact restoration of the baseline activity after transient responses to step stimuli (Barkai and Leibler, 1997; Yi et al, 2000). However, the output of time-derivative computations during smoothly varying stimuli depends not only on the presence of integral feedback, but also on what is being integrated. As this integrand can in general be any function of the output, we represent it by a black-box function F(a) in our model, and set out to determine its shape by experiments with time-varying stimuli.
We first apply exponential ramp stimuli—waveforms in which the logarithm of the stimulus level varies linearly with time, at a fixed rate r. It was shown many years ago that during such a stimulus, the kinase output of the pathway changes to a new constant value, ac that is dependent on the applied ramp rate, r (Block et al, 1983). A plot of ac versus r (Figure 5A) can thus be considered as an output of time-derivative computations by the network, and could also be used to study the ‘gradient sensitivity' of bacteria traveling at constant speeds.
To obtain the feedback transfer function, F(a), we apply a simple coordinate transformation, identified using our model, to the same ramp-response data (Figure 5B). This function reveals how the temporal rate of change of the feedback signal m depends on the current output signal a. The shape of this function is analyzed using a biochemical reaction scheme, from which in vivo kinetic parameters of the feedback enzymes, CheR and CheB, are extracted. The fitted Michaelis constants for these enzymatic reactions are small compared with the steady-state abundance of their substrates, thus indicating that these enzymes operate close to saturation in vivo. The slope of the function near steady state can be used to assess the strength of feedback, and to compute the relaxation time of the system, τm. Relaxation is found to be slow (i.e. large τm), consistent with large fluctuations about the steady-state activity caused by the near-saturation kinetics of the feedback enzymes (Emonet and Cluzel, 2008).
Finally, exponential sine-wave stimuli are used to map out the system's frequency response (Figure 5C). The measured data points for both the amplitude and phase of the response are found to be in excellent agreement with model predictions based on parameters from the independently measured step and ramp responses. No curve fitting was required to obtain this agreement. Although the amplitude response as a function of frequency resembles a first-order high-pass filter with a well-defined cutoff frequency, νm, we point out that the chemotaxis pathway is actually a low-pass filter if the time derivative of the input is viewed as the input signal. In this latter perspective, νm defines an upper bound for the frequency band over which time-derivative computations can be carried out.
The two types of measurements yield complementary information regarding time-derivative computations by E. coli. The ramp-responses characterize the asymptotically constant output when a temporal gradient is held fixed over extended periods. Interestingly, the ramp responses do not depend on receptor cooperativity, but only on properties of the adaptation system, and thus can be used to reveal the in vivo adaptation kinetics, even outside the linear regime of the kinase response. The frequency response is highly relevant in considering spatial searches in the real world, in which experienced gradients are not held fixed in time. The characteristic cutoff frequency νm is found by working within the linear regime of the kinase response, and depends on parameters from both modules (it increases with both cooperativity in the receptor module, and the strength of feedback in the adaptation module).
Both ramp responses and sine-wave responses were measured at two different temperatures (22 and 32°C), and found to differ significantly. Both the slope of F(a) near steady state, from ramp experiments, and the characteristic cutoff frequency, from sine-wave experiments, were higher by a factor of ∼3 at 32°C. Fits of the enzymatic model to F(a) suggest that temperature affects the maximal velocity (Vmax) more strongly than the Michaelis constants (Km) for CheR and CheB.
Successful application of inter-molecular FRET in live cells using GFP variants always requires some degree of serendipity. Genetic fusions to these bulky fluorophores can impair the function of the original proteins, and even when fusions are functional, efficient FRET still requires the fused fluorophores to come within the small (<10 nm) Förster radius on interactions between the labeled proteins. Thus, when a successful FRET pair is identified, it is desirable to make the most of it. We have shown here that combined with careful temporal control of input stimuli, and appropriately calibrated models, a single FRET pair can be used to study the structure of multiple transfer functions within a signaling network.
The Escherichia coli chemotaxis-signaling pathway computes time derivatives of chemoeffector concentrations. This network features modules for signal reception/amplification and robust adaptation, with sensing of chemoeffector gradients determined by the way in which these modules are coupled in vivo. We characterized these modules and their coupling by using fluorescence resonance energy transfer to measure intracellular responses to time-varying stimuli. Receptor sensitivity was characterized by step stimuli, the gradient sensitivity by exponential ramp stimuli, and the frequency response by exponential sine-wave stimuli. Analysis of these data revealed the structure of the feedback transfer function linking the amplification and adaptation modules. Feedback near steady state was found to be weak, consistent with strong fluctuations and slow recovery from small perturbations. Gradient sensitivity and frequency response both depended strongly on temperature. We found that time derivatives can be computed by the chemotaxis system for input frequencies below 0.006 Hz at 22°C and below 0.018 Hz at 32°C. Our results show how dynamic input–output measurements, time honored in physiology, can serve as powerful tools in deciphering cell-signaling mechanisms.
PMCID: PMC2913400  PMID: 20571531
adaptation; feedback; fluorescence resonance energy transfer (FRET); frequency response; Monod–Wyman–Changeux (MWC) model
11.  ScreenMill: A freely available software suite for growth measurement, analysis and visualization of high-throughput screen data 
BMC Bioinformatics  2010;11:353.
Many high-throughput genomic experiments, such as Synthetic Genetic Array and yeast two-hybrid, use colony growth on solid media as a screen metric. These experiments routinely generate over 100,000 data points, making data analysis a time consuming and painstaking process. Here we describe ScreenMill, a new software suite that automates image analysis and simplifies data review and analysis for high-throughput biological experiments.
The ScreenMill, software suite includes three software tools or "engines": an open source Colony Measurement Engine (CM Engine) to quantitate colony growth data from plate images, a web-based Data Review Engine (DR Engine) to validate and analyze quantitative screen data, and a web-based Statistics Visualization Engine (SV Engine) to visualize screen data with statistical information overlaid. The methods and software described here can be applied to any screen in which growth is measured by colony size. In addition, the DR Engine and SV Engine can be used to visualize and analyze other types of quantitative high-throughput data.
ScreenMill automates quantification, analysis and visualization of high-throughput screen data. The algorithms implemented in ScreenMill are transparent allowing users to be confident about the results ScreenMill produces. Taken together, the tools of ScreenMill offer biologists a simple and flexible way of analyzing their data, without requiring programming skills.
PMCID: PMC2909220  PMID: 20584323
12.  Single-Cell Tracking Reveals Antibiotic-Induced Changes in Mycobacterial Energy Metabolism 
mBio  2015;6(1):e02236-14.
ATP is a key molecule of cell physiology, but despite its importance, there are currently no methods for monitoring single-cell ATP fluctuations in live bacteria. This is a major obstacle in studies of bacterial energy metabolism, because there is a growing awareness that bacteria respond to stressors such as antibiotics in a highly individualistic manner. Here, we present a method for long-term single-cell tracking of ATP levels in Mycobacterium smegmatis based on a combination of microfluidics, time-lapse microscopy, and Förster resonance energy transfer (FRET)-based ATP biosensors. Upon treating cells with antibiotics, we observed that individual cells undergo an abrupt and irreversible switch from high to low intracellular ATP levels. The kinetics and extent of ATP switching clearly discriminate between an inhibitor of ATP synthesis and other classes of antibiotics. Cells that resume growth after 24 h of antibiotic treatment maintain high ATP levels throughout the exposure period. In contrast, antibiotic-treated cells that switch from ATP-high to ATP-low states never resume growth after antibiotic washout. Surprisingly, only a subset of these nongrowing ATP-low cells stains with propidium iodide (PI), a widely used live/dead cell marker. These experiments also reveal a cryptic subset of cells that do not resume growth after antibiotic washout despite remaining ATP high and PI negative. We conclude that ATP tracking is a more dynamic, sensitive, reliable, and discriminating marker of cell viability than staining with PI. This method could be used in studies to evaluate antimicrobial effectiveness and mechanism of action, as well as for high-throughput screening.
New antimicrobials are urgently needed to stem the rising tide of antibiotic-resistant bacteria. All antibiotics are expected to affect bacterial energy metabolism, directly or indirectly, yet tools to assess the impact of antibiotics on the ATP content of individual bacterial cells are lacking. The method described here for single-cell tracking of intracellular ATP in live bacteria has many advantages compared to conventional ensemble-averaged assays. It provides a continuous real-time readout of bacterial ATP content, cell vitality, and antimicrobial mechanism of action with high temporal resolution at the single-cell level. In combination with high-throughput microfluidic devices and automated microscopy, this method also has the potential to serve as a novel screening tool in antimicrobial drug discovery.
PMCID: PMC4338811  PMID: 25691591
13.  Analysis of in vivo single cell behavior by high throughput, human-in-the-loop segmentation of three-dimensional images 
BMC Bioinformatics  2015;16:397.
Analysis of single cells in their native environment is a powerful method to address key questions in developmental systems biology. Confocal microscopy imaging of intact tissues, followed by automatic image segmentation, provides a means to conduct cytometric studies while at the same time preserving crucial information about the spatial organization of the tissue and morphological features of the cells. This technique is rapidly evolving but is still not in widespread use among research groups that do not specialize in technique development, perhaps in part for lack of tools that automate repetitive tasks while allowing experts to make the best use of their time in injecting their domain-specific knowledge.
Here we focus on a well-established stem cell model system, the C. elegans gonad, as well as on two other model systems widely used to study cell fate specification and morphogenesis: the pre-implantation mouse embryo and the developing mouse olfactory epithelium. We report a pipeline that integrates machine-learning-based cell detection, fast human-in-the-loop curation of these detections, and running of active contours seeded from detections to segment cells. The procedure can be bootstrapped by a small number of manual detections, and outperforms alternative pieces of software we benchmarked on C. elegans gonad datasets. Using cell segmentations to quantify fluorescence contents, we report previously-uncharacterized cell behaviors in the model systems we used. We further show how cell morphological features can be used to identify cell cycle phase; this provides a basis for future tools that will streamline cell cycle experiments by minimizing the need for exogenous cell cycle phase labels.
High-throughput 3D segmentation makes it possible to extract rich information from images that are routinely acquired by biologists, and provides insights — in particular with respect to the cell cycle — that would be difficult to derive otherwise.
Electronic supplementary material
The online version of this article (doi:10.1186/s12859-015-0814-7) contains supplementary material, which is available to authorized users.
PMCID: PMC4659165  PMID: 26607933
Spatial cytometry; 3D image segmentation; Stem cells; Cell cycle; C. elegans germ line; Mouse pre-implantation embryo; Olfactory placode; Olfactory epithelium
14.  An image score inference system for RNAi genome-wide screening based on fuzzy mixture regression modeling 
With recent advances in fluorescence microscopy imaging techniques and methods of gene knock down by RNA interference (RNAi), genome-scale high-content screening (HCS) has emerged as a powerful approach to systematically identify all parts of complex biological processes. However, a critical barrier preventing fulfillment of the success is the lack of efficient and robust methods for automating RNAi image analysis and quantitative evaluation of the gene knock down effects on huge volume of HCS data. Facing such opportunities and challenges, we have started investigation of automatic methods towards the development of a fully automatic RNAi-HCS system. Particularly important are reliable approaches to cellular phenotype classification and image-based gene function estimation.
We have developed a HCS analysis platform that consists of two main components: fluorescence image analysis and image scoring. For image analysis, we used a two-step enhanced watershed method to extract cellular boundaries from HCS images. Segmented cells were classified into several predefined phenotypes based on morphological and appearance features. Using statistical characteristics of the identified phenotypes as a quantitative description of the image, a score is generated that reflects gene function. Our scoring model integrates fuzzy gene class estimation and single regression models. The final functional score of an image was derived using the weighted combination of the inference from several support vector-based regression models. We validated our phenotype classification method and scoring system on our cellular phenotype and gene database with expert ground truth labeling.
We built a database of high-content, 3-channel, fluorescence microscopy images of Drosophila Kc167 cultured cells that were treated with RNAi to perturb gene function. The proposed informatics system for microscopy image analysis is tested on this database. Both of the two main components, automated phenotype classification and image scoring system, were evaluated. The robustness and efficiency of our system were validated in quantitatively predicting the biological relevance of genes.
PMCID: PMC2763194  PMID: 18547870
High-content screening; Image score inference
15.  Shedding Light on Filovirus Infection with High-Content Imaging 
Viruses  2012;4(8):1354-1371.
Microscopy has been instrumental in the discovery and characterization of microorganisms. Major advances in high-throughput fluorescence microscopy and automated, high-content image analysis tools are paving the way to the systematic and quantitative study of the molecular properties of cellular systems, both at the population and at the single-cell level. High-Content Imaging (HCI) has been used to characterize host-virus interactions in genome-wide reverse genetic screens and to identify novel cellular factors implicated in the binding, entry, replication and egress of several pathogenic viruses. Here we present an overview of the most significant applications of HCI in the context of the cell biology of filovirus infection. HCI assays have been recently implemented to quantitatively study filoviruses in cell culture, employing either infectious viruses in a BSL-4 environment or surrogate genetic systems in a BSL-2 environment. These assays are becoming instrumental for small molecule and siRNA screens aimed at the discovery of both cellular therapeutic targets and of compounds with anti-viral properties. We discuss the current practical constraints limiting the implementation of high-throughput biology in a BSL-4 environment, and propose possible solutions to safely perform high-content, high-throughput filovirus infection assays. Finally, we discuss possible novel applications of HCI in the context of filovirus research with particular emphasis on the identification of possible cellular biomarkers of virus infection.
PMCID: PMC3446768  PMID: 23012631
filoviruses; High-Content Imaging; therapeutics; host-pathogen interactions; phenotype
16.  Deep Learning Automates the Quantitative Analysis of Individual Cells in Live-Cell Imaging Experiments 
PLoS Computational Biology  2016;12(11):e1005177.
Live-cell imaging has opened an exciting window into the role cellular heterogeneity plays in dynamic, living systems. A major critical challenge for this class of experiments is the problem of image segmentation, or determining which parts of a microscope image correspond to which individual cells. Current approaches require many hours of manual curation and depend on approaches that are difficult to share between labs. They are also unable to robustly segment the cytoplasms of mammalian cells. Here, we show that deep convolutional neural networks, a supervised machine learning method, can solve this challenge for multiple cell types across the domains of life. We demonstrate that this approach can robustly segment fluorescent images of cell nuclei as well as phase images of the cytoplasms of individual bacterial and mammalian cells from phase contrast images without the need for a fluorescent cytoplasmic marker. These networks also enable the simultaneous segmentation and identification of different mammalian cell types grown in co-culture. A quantitative comparison with prior methods demonstrates that convolutional neural networks have improved accuracy and lead to a significant reduction in curation time. We relay our experience in designing and optimizing deep convolutional neural networks for this task and outline several design rules that we found led to robust performance. We conclude that deep convolutional neural networks are an accurate method that require less curation time, are generalizable to a multiplicity of cell types, from bacteria to mammalian cells, and expand live-cell imaging capabilities to include multi-cell type systems.
Author Summary
Dynamic live-cell imaging experiments are a powerful tool to interrogate biological systems with single cell resolution. The key barrier to analyzing data generated by these measurements is image segmentation—identifying which parts of an image belong to which individual cells. Here we show that deep learning is a natural technology to solve this problem for these experiments. We show that deep learning is more accurate, requires less time to curate segmentation results, can segment multiple cell types, and can distinguish between different cell lines present in the same image. We highlight specific design rules that enable us to achieve high segmentation accuracy even with a small number of manually annotated images (~100 cells). We expect that our work will enable new experiments that were previously impossible, as well as reduce the computational barrier for new labs to join the live-cell imaging space.
PMCID: PMC5096676  PMID: 27814364
17.  CellSegm - a MATLAB toolbox for high-throughput 3D cell segmentation 
The application of fluorescence microscopy in cell biology often generates a huge amount of imaging data. Automated whole cell segmentation of such data enables the detection and analysis of individual cells, where a manual delineation is often time consuming, or practically not feasible. Furthermore, compared to manual analysis, automation normally has a higher degree of reproducibility. CellSegm, the software presented in this work, is a Matlab based command line software toolbox providing an automated whole cell segmentation of images showing surface stained cells, acquired by fluorescence microscopy. It has options for both fully automated and semi-automated cell segmentation. Major algorithmic steps are: (i) smoothing, (ii) Hessian-based ridge enhancement, (iii) marker-controlled watershed segmentation, and (iv) feature-based classfication of cell candidates. Using a wide selection of image recordings and code snippets, we demonstrate that CellSegm has the ability to detect various types of surface stained cells in 3D. After detection and outlining of individual cells, the cell candidates can be subject to software based analysis, specified and programmed by the end-user, or they can be analyzed by other software tools. A segmentation of tissue samples with appropriate characteristics is also shown to be resolvable in CellSegm. The command-line interface of CellSegm facilitates scripting of the separate tools, all implemented in Matlab, offering a high degree of flexibility and tailored workflows for the end-user. The modularity and scripting capabilities of CellSegm enable automated workflows and quantitative analysis of microscopic data, suited for high-throughput image based screening.
PMCID: PMC3850890  PMID: 23938087
Automated analysis; Cell segmentation; CellSegm; High-throughput; Nucleus staining; Surface staining
18.  Comparison of Microscopy and Alamar Blue Reduction in a Larval Based Assay for Schistosome Drug Screening 
In view of the current widespread use of and reliance on a single schistosomicide, praziquantel, there is a pressing need to discover and develop alternative drugs for schistosomiasis. One approach to this is to develop High Throughput in vitro whole organism screens (HTS) to identify hits amongst large compound libraries.
Methodology/Principal Findings
We have been carrying out low throughput (24-well plate) in vitro testing based on microscopic evaluation of killing of ex-vivo adult S. mansoni worms using selected compound collections mainly provided through the WHO-TDR Helminth Drug Initiative. To increase throughput, we introduced a similar but higher throughput 96-well primary in vitro assay using the schistosomula stage which can be readily produced in vitro in large quantities. In addition to morphological readout of viability we have investigated using fluorometric determination of the reduction of Alamar blue (AB), a redox indicator of enzyme activity widely used in whole organism screening. A panel of 7 known schistosome active compounds including praziquantel, produced diverse effects on larval morphology within 3 days of culture although only two induced marked larval death within 7 days. The AB assay was very effective in detecting these lethal compounds but proved more inconsistent in detecting compounds which damaged but did not kill. The utility of the AB assay in detecting compounds which cause severe morbidity and/or death of schistosomula was confirmed in testing a panel of compounds previously selected in library screening as having activity against the adult worms. Furthermore, in prospective library screening, the AB assay was able to detect all compounds which induced killing and also the majority of compounds designated as hits based on morphological changes.
We conclude that an HTS combining AB readout and image-based analysis would provide an efficient and stringent primary assay for schistosome drug discovery.
Author Summary
Only one drug, praziquantel, is widely available for treating schistosomiasis, a disease affecting an estimated 200 million people. Because of increasing usage there is concern about development of praziquantel drug resistance and a perceived need to develop new schistosomicides. Possible sources of these are large collections of compounds held by pharmaceutical companies and academic institutions. Anti-schistosome activity can be detected in vitro by visually assessing damage to cultured adult schistosome worms, but these are large and are recovered from mice which somewhat limits screening throughput. By contrast, schistosomula can be produced in vitro and used for screening in microwell plates, thus allowing medium throughput screening. High throughput screening (HTS) would require automated readout of schistosomulicidal action rather than manual microscopy. Here we report on the use of Alamar blue (AB), a fluorescent indicator of cell viability which can be measured rapidly and automatically. The AB assay was readily able to detect compounds causing death or severe damage to the larvae but was less reliable than microscopy for more subtle morphological changes including those induced by some known schistosome drugs. It is concluded that an automated HTS would benefit from integrated use of both AB and automatic image-based morphology assays.
PMCID: PMC2919390  PMID: 20706580
19.  Optimal Experimental Design for Parameter Estimation of a Cell Signaling Model 
PLoS Computational Biology  2009;5(11):e1000558.
Differential equation models that describe the dynamic changes of biochemical signaling states are important tools to understand cellular behavior. An essential task in building such representations is to infer the affinities, rate constants, and other parameters of a model from actual measurement data. However, intuitive measurement protocols often fail to generate data that restrict the range of possible parameter values. Here we utilized a numerical method to iteratively design optimal live-cell fluorescence microscopy experiments in order to reveal pharmacological and kinetic parameters of a phosphatidylinositol 3,4,5-trisphosphate (PIP3) second messenger signaling process that is deregulated in many tumors. The experimental approach included the activation of endogenous phosphoinositide 3-kinase (PI3K) by chemically induced recruitment of a regulatory peptide, reversible inhibition of PI3K using a kinase inhibitor, and monitoring of the PI3K-mediated production of PIP3 lipids using the pleckstrin homology (PH) domain of Akt. We found that an intuitively planned and established experimental protocol did not yield data from which relevant parameters could be inferred. Starting from a set of poorly defined model parameters derived from the intuitively planned experiment, we calculated concentration-time profiles for both the inducing and the inhibitory compound that would minimize the predicted uncertainty of parameter estimates. Two cycles of optimization and experimentation were sufficient to narrowly confine the model parameters, with the mean variance of estimates dropping more than sixty-fold. Thus, optimal experimental design proved to be a powerful strategy to minimize the number of experiments needed to infer biological parameters from a cell signaling assay.
Author Summary
Differential equation models of signaling processes are useful to gain a molecular and quantitative understanding of cellular information flow. Although these models are typically based on simple kinetic rules, they can often qualitatively describe the behavior of biological systems. However, in the quest to transform biomedical research into an engineering discipline, biologists face the challenge of estimating important parameters of such models from laboratory data. Measurement noise as well as the robust architecture of biological circuits are causes for large uncertainty of parameter estimates. This makes it difficult to plan informative experiments. Here, we used a computational method to predict and minimize the uncertainty of parameter estimates we would obtain from prospective experiments given a cancer-relevant signaling model. This was achieved by optimizing the concentrations and time points for adding drugs in a live-cell microscopy experiment. Our experimental results demonstrated that the advice given by this algorithm resulted in many-fold more informative data than we would obtain by repeating an intuitively planned experiment. Our study shows that significant experimental effort and time can be saved by adopting an optimal experimental design strategy for inferring relevant parameters from biomedical experiments.
PMCID: PMC2775273  PMID: 19911077
20.  Unsupervised Clustering of Subcellular Protein Expression Patterns in High-Throughput Microscopy Images Reveals Protein Complexes and Functional Relationships between Proteins 
PLoS Computational Biology  2013;9(6):e1003085.
Protein subcellular localization has been systematically characterized in budding yeast using fluorescently tagged proteins. Based on the fluorescence microscopy images, subcellular localization of many proteins can be classified automatically using supervised machine learning approaches that have been trained to recognize predefined image classes based on statistical features. Here, we present an unsupervised analysis of protein expression patterns in a set of high-resolution, high-throughput microscope images. Our analysis is based on 7 biologically interpretable features which are evaluated on automatically identified cells, and whose cell-stage dependency is captured by a continuous model for cell growth. We show that it is possible to identify most previously identified localization patterns in a cluster analysis based on these features and that similarities between the inferred expression patterns contain more information about protein function than can be explained by a previous manual categorization of subcellular localization. Furthermore, the inferred cell-stage associated to each fluorescence measurement allows us to visualize large groups of proteins entering the bud at specific stages of bud growth. These correspond to proteins localized to organelles, revealing that the organelles must be entering the bud in a stereotypical order. We also identify and organize a smaller group of proteins that show subtle differences in the way they move around the bud during growth. Our results suggest that biologically interpretable features based on explicit models of cell morphology will yield unprecedented power for pattern discovery in high-resolution, high-throughput microscopy images.
Author Summary
The location of a particular protein in the cell is one of the most important pieces of information that cell biologists use to understand its function. Fluorescent tags are a powerful way to determine the location of a protein in living cells. Nearly a decade ago, a collection of yeast strains was introduced, where in each strain a single protein was tagged with green fluorescent protein (GFP). Here, we show that by training a computer to accurately identify the buds of growing yeast cells, and then making simple fluorescence measurements in context of cell shape and cell stage, the computer could automatically discover most of the localization patterns (nucleus, cytoplasm, mitochondria, etc.) without any prior knowledge of what the patterns might be. Because we made the same, simple measurements for each yeast cell, we could compare and visualize the patterns of fluorescence for the entire collection of strains. This allowed us to identify large groups of proteins moving around the cell in a coordinated fashion, and to identify new, complex patterns that had previously been difficult to describe.
PMCID: PMC3681667  PMID: 23785265
21.  Visualization and correction of automated segmentation, tracking and lineaging from 5-D stem cell image sequences 
BMC Bioinformatics  2014;15(1):328.
Neural stem cells are motile and proliferative cells that undergo mitosis, dividing to produce daughter cells and ultimately generating differentiated neurons and glia. Understanding the mechanisms controlling neural stem cell proliferation and differentiation will play a key role in the emerging fields of regenerative medicine and cancer therapeutics. Stem cell studies in vitro from 2-D image data are well established. Visualizing and analyzing large three dimensional images of intact tissue is a challenging task. It becomes more difficult as the dimensionality of the image data increases to include time and additional fluorescence channels. There is a pressing need for 5-D image analysis and visualization tools to study cellular dynamics in the intact niche and to quantify the role that environmental factors play in determining cell fate.
We present an application that integrates visualization and quantitative analysis of 5-D (x,y,z,t,channel) and large montage confocal fluorescence microscopy images. The image sequences show stem cells together with blood vessels, enabling quantification of the dynamic behaviors of stem cells in relation to their vascular niche, with applications in developmental and cancer biology. Our application automatically segments, tracks, and lineages the image sequence data and then allows the user to view and edit the results of automated algorithms in a stereoscopic 3-D window while simultaneously viewing the stem cell lineage tree in a 2-D window. Using the GPU to store and render the image sequence data enables a hybrid computational approach. An inference-based approach utilizing user-provided edits to automatically correct related mistakes executes interactively on the system CPU while the GPU handles 3-D visualization tasks.
By exploiting commodity computer gaming hardware, we have developed an application that can be run in the laboratory to facilitate rapid iteration through biological experiments. We combine unsupervised image analysis algorithms with an interactive visualization of the results. Our validation interface allows for each data set to be corrected to 100% accuracy, ensuring that downstream data analysis is accurate and verifiable. Our tool is the first to combine all of these aspects, leveraging the synergies obtained by utilizing validation information from stereo visualization to improve the low level image processing tasks.
Electronic supplementary material
The online version of this article (doi:10.1186/1471-2105-15-328) contains supplementary material, which is available to authorized users.
PMCID: PMC4287543  PMID: 25281197
Volumetric texture rendering; 3-D display; Stereoscopic 3-D; Stem cell; Time lapse; Lineaging; Validation and correction; Confocal microscopy; CUDA; Image montage reconstruction
22.  Computer-based fluorescence quantification: a novel approach to study nucleolar biology 
BMC Cell Biology  2011;12:25.
Nucleoli are composed of possibly several thousand different proteins and represent the most conspicuous compartments in the nucleus; they play a crucial role in the proper execution of many cellular processes. As such, nucleoli carry out ribosome biogenesis and sequester or associate with key molecules that regulate cell cycle progression, tumorigenesis, apoptosis and the stress response. Nucleoli are dynamic compartments that are characterized by a constant flux of macromolecules. Given the complex and dynamic composition of the nucleolar proteome, it is challenging to link modifications in nucleolar composition to downstream effects.
In this contribution, we present quantitative immunofluorescence methods that rely on computer-based image analysis. We demonstrate the effectiveness of these techniques by monitoring the dynamic association of proteins and RNA with nucleoli under different physiological conditions. Thus, the protocols described by us were employed to study stress-dependent changes in the nucleolar concentration of endogenous and GFP-tagged proteins. Furthermore, our methods were applied to measure de novo RNA synthesis that is associated with nucleoli. We show that the techniques described here can be easily combined with automated high throughput screening (HTS) platforms, making it possible to obtain large data sets and analyze many of the biological processes that are located in nucleoli.
Our protocols set the stage to analyze in a quantitative fashion the kinetics of shuttling nucleolar proteins, both at the single cell level as well as for a large number of cells. Moreover, the procedures described here are compatible with high throughput image acquisition and analysis using HTS automated platforms, thereby providing the basis to quantify nucleolar components and activities for numerous samples and experimental conditions. Together with the growing amount of information obtained for the nucleolar proteome, improvements in quantitative microscopy as they are described here can be expected to produce new insights into the complex biological functions that are orchestrated by the nucleolus.
PMCID: PMC3126779  PMID: 21639891
23.  Rapid Analysis and Exploration of Fluorescence Microscopy Images 
Despite rapid advances in high-throughput microscopy, quantitative image-based assays still pose significant challenges. While a variety of specialized image analysis tools are available, most traditional image-analysis-based workflows have steep learning curves (for fine tuning of analysis parameters) and result in long turnaround times between imaging and analysis. In particular, cell segmentation, the process of identifying individual cells in an image, is a major bottleneck in this regard.
Here we present an alternate, cell-segmentation-free workflow based on PhenoRipper, an open-source software platform designed for the rapid analysis and exploration of microscopy images. The pipeline presented here is optimized for immunofluorescence microscopy images of cell cultures and requires minimal user intervention. Within half an hour, PhenoRipper can analyze data from a typical 96-well experiment and generate image profiles. Users can then visually explore their data, perform quality control on their experiment, ensure response to perturbations and check reproducibility of replicates. This facilitates a rapid feedback cycle between analysis and experiment, which is crucial during assay optimization. This protocol is useful not just as a first pass analysis for quality control, but also may be used as an end-to-end solution, especially for screening. The workflow described here scales to large data sets such as those generated by high-throughput screens, and has been shown to group experimental conditions by phenotype accurately over a wide range of biological systems. The PhenoBrowser interface provides an intuitive framework to explore the phenotypic space and relate image properties to biological annotations. Taken together, the protocol described here will lower the barriers to adopting quantitative analysis of image based screens.
PMCID: PMC4390293  PMID: 24686220
PhenoRipper; Fluorescence Microscopy; Image Analysis; High-content analysis; High-throughput screening; Open-source; Phenotype
24.  Pipeline for illumination correction of images for high-throughput microscopy 
Journal of Microscopy  2014;256(3):231-236.
The presence of systematic noise in images in high-throughput microscopy experiments can significantly impact the accuracy of downstream results. Among the most common sources of systematic noise is non-homogeneous illumination across the image field. This often adds an unacceptable level of noise, obscures true quantitative differences and precludes biological experiments that rely on accurate fluorescence intensity measurements.
In this paper, we seek to quantify the improvement in the quality of high-content screen readouts due to software-based illumination correction. We present a straightforward illumination correction pipeline that has been used by our group across many experiments. We test the pipeline on real-world high-throughput image sets and evaluate the performance of the pipeline at two levels: (a) Z′-factor to evaluate the effect of the image correction on a univariate readout, representative of a typical high-content screen, and (b) classification accuracy on phenotypic signatures derived from the images, representative of an experiment involving more complex data mining. We find that applying the proposed post-hoc correction method improves performance in both experiments, even when illumination correction has already been applied using software associated with the instrument.
To facilitate the ready application and future development of illumination correction methods, we have made our complete test data sets as well as open-source image analysis pipelines publicly available. This software-based solution has the potential to improve outcomes for a wide-variety of image-based HTS experiments.
PMCID: PMC4359755  PMID: 25228240
Fluorescence microscopy; high-throughput microscopy; illumination correction; shading correction; vignetting
25.  Developing an efficient scheduling template of a chemotherapy treatment unit 
The Australasian Medical Journal  2011;4(10):575-588.
This study was undertaken to improve the performance of a Chemotherapy Treatment Unit by increasing the throughput and reducing the average patient’s waiting time. In order to achieve this objective, a scheduling template has been built. The scheduling template is a simple tool that can be used to schedule patients' arrival to the clinic. A simulation model of this system was built and several scenarios, that target match the arrival pattern of the patients and resources availability, were designed and evaluated. After performing detailed analysis, one scenario provide the best system’s performance. A scheduling template has been developed based on this scenario. After implementing the new scheduling template, 22.5% more patients can be served.
CancerCare Manitoba is a provincially mandated cancer care agency. It is dedicated to provide quality care to those who have been diagnosed and are living with cancer. MacCharles Chemotherapy unit is specially built to provide chemotherapy treatment to the cancer patients of Winnipeg. In order to maintain an excellent service, it tries to ensure that patients get their treatment in a timely manner. It is challenging to maintain that goal because of the lack of a proper roster, the workload distribution and inefficient resource allotment. In order to maintain the satisfaction of the patients and the healthcare providers, by serving the maximum number of patients in a timely manner, it is necessary to develop an efficient scheduling template that matches the required demand with the availability of resources. This goal can be reached using simulation modelling. Simulation has proven to be an excellent modelling tool. It can be defined as building computer models that represent real world or hypothetical systems, and hence experimenting with these models to study system behaviour under different scenarios.1, 2
A study was undertaken at the Children's Hospital of Eastern Ontario to identify the issues behind the long waiting time of a emergency room.3 A 20-­‐day field observation revealed that the availability of the staff physician and interaction affects the patient wait time. Jyväskylä et al.4 used simulation to test different process scenarios, allocate resources and perform activity-­‐based cost analysis in the Emergency Department (ED) at the Central Hospital. The simulation also supported the study of a new operational method, named "triage-team" method without interrupting the main system. The proposed triage team method categorises the entire patient according to the urgency to see the doctor and allows the patient to complete the necessary test before being seen by the doctor for the first time. The simulation study showed that it will decrease the throughput time of the patient and reduce the utilisation of the specialist and enable the ordering all the tests the patient needs right after arrival, thus quickening the referral to treatment.
Santibáñez et al.5 developed a discrete event simulation model of British Columbia Cancer Agency"s ambulatory care unit which was used to study the impact of scenarios considering different operational factors (delay in starting clinic), appointment schedule (appointment order, appointment adjustment, add-­‐ons to the schedule) and resource allocation. It was found that the best outcomes were obtained when not one but multiple changes were implemented simultaneously. Sepúlveda et al.6 studied the M. D. Anderson Cancer Centre Orlando, which is a cancer treatment facility and built a simulation model to analyse and improve flow process and increase capacity in the main facility. Different scenarios were considered like, transferring laboratory and pharmacy areas, adding an extra blood draw room and applying different scheduling techniques of patients. The study shows that by increasing the number of short-­‐term (four hours or less) patients in the morning could increase chair utilisation.
Discrete event simulation also helps improve a service where staff are ignorant about the behaviour of the system as a whole; which can also be described as a real professional system. Niranjon et al.7 used simulation successfully where they had to face such constraints and lack of accessible data. Carlos et al. 8 used Total quality management and simulation – animation to improve the quality of the emergency room. Simulation was used to cover the key point of the emergency room and animation was used to indicate the areas of opportunity required. This study revealed that a long waiting time, overload personnel and increasing withdrawal rate of patients are caused by the lack of capacity in the emergency room.
Baesler et al.9 developed a methodology for a cancer treatment facility to find stochastically a global optimum point for the control variables. A simulation model generated the output using a goal programming framework for all the objectives involved in the analysis. Later a genetic algorithm was responsible for performing the search for an improved solution. The control variables that were considered in this research are number of treatment chairs, number of drawing blood nurses, laboratory personnel, and pharmacy personnel. Guo et al. 10 presented a simulation framework considering demand for appointment, patient flow logic, distribution of resources, scheduling rules followed by the scheduler. The objective of the study was to develop a scheduling rule which will ensure that 95% of all the appointment requests should be seen within one week after the request is made to increase the level of patient satisfaction and balance the schedule of each doctor to maintain a fine harmony between "busy clinic" and "quiet clinic".
Huschka et al.11 studied a healthcare system which was about to change their facility layout. In this case a simulation model study helped them to design a new healthcare practice by evaluating the change in layout before implementation. Historical data like the arrival rate of the patients, number of patients visited each day, patient flow logic, was used to build the current system model. Later, different scenarios were designed which measured the changes in the current layout and performance.
Wijewickrama et al.12 developed a simulation model to evaluate appointment schedule (AS) for second time consultations and patient appointment sequence (PSEQ) in a multi-­‐facility system. Five different appointment rule (ARULE) were considered: i) Baily; ii) 3Baily; iii) Individual (Ind); iv) two patients at a time (2AtaTime); v) Variable Interval and (V-­‐I) rule. PSEQ is based on type of patients: Appointment patients (APs) and new patients (NPs). The different PSEQ that were studied in this study were: i) first-­‐ come first-­‐serve; ii) appointment patient at the beginning of the clinic (APBEG); iii) new patient at the beginning of the clinic (NPBEG); iv) assigning appointed and new patients in an alternating manner (ALTER); v) assigning a new patient after every five-­‐appointment patients. Also patient no show (0% and 5%) and patient punctuality (PUNCT) (on-­‐time and 10 minutes early) were also considered. The study found that ALTER-­‐Ind. and ALTER5-­‐Ind. performed best on 0% NOSHOW, on-­‐time PUNCT and 5% NOSHOW, on-­‐time PUNCT situation to reduce WT and IT per patient. As NOSHOW created slack time for waiting patients, their WT tends to reduce while IT increases due to unexpected cancellation. Earliness increases congestion whichin turn increases waiting time.
Ramis et al.13 conducted a study of a Medical Imaging Center (MIC) to build a simulation model which was used to improve the patient journey through an imaging centre by reducing the wait time and making better use of the resources. The simulation model also used a Graphic User Interface (GUI) to provide the parameters of the centre, such as arrival rates, distances, processing times, resources and schedule. The simulation was used to measure the waiting time of the patients in different case scenarios. The study found that assigning a common function to the resource personnel could improve the waiting time of the patients.
The objective of this study is to develop an efficient scheduling template that maximises the number of served patients and minimises the average patient's waiting time at the given resources availability. To accomplish this objective, we will build a simulation model which mimics the working conditions of the clinic. Then we will suggest different scenarios of matching the arrival pattern of the patients with the availability of the resources. Full experiments will be performed to evaluate these scenarios. Hence, a simple and practical scheduling template will be built based on the indentified best scenario. The developed simulation model is described in section 2, which consists of a description of the treatment room, and a description of the types of patients and treatment durations. In section 3, different improvement scenarios are described and their analysis is presented in section 4. Section 5 illustrates a scheduling template based on one of the improvement scenarios. Finally, the conclusion and future direction of our work is exhibited in section 6.
Simulation Model
A simulation model represents the actual system and assists in visualising and evaluating the performance of the system under different scenarios without interrupting the actual system. Building a proper simulation model of a system consists of the following steps.
Observing the system to understand the flow of the entities, key players, availability of resources and overall generic framework.
Collecting the data on the number and type of entities, time consumed by the entities at each step of their journey, and availability of resources.
After building the simulation model it is necessary to confirm that the model is valid. This can be done by confirming that each entity flows as it is supposed to and the statistical data generated by the simulation model is similar to the collected data.
Figure 1 shows the patient flow process in the treatment room. On the patient's first appointment, the oncologist comes up with the treatment plan. The treatment time varies according to the patient’s condition, which may be 1 hour to 10 hours. Based on the type of the treatment, the physician or the clinical clerk books an available treatment chair for that time period.
On the day of the appointment, the patient will wait until the booked chair is free. When the chair is free a nurse from that station comes to the patient, verifies the name and date of birth and takes the patient to a treatment chair. Afterwards, the nurse flushes the chemotherapy drug line to the patient's body which takes about five minutes and sets up the treatment. Then the nurse leaves to serve another patient. Chemotherapy treatment lengths vary from less than an hour to 10 hour infusions. At the end of the treatment, the nurse returns, removes the line and notifies the patient about the next appointment date and time which also takes about five minutes. Most of the patients visit the clinic to take care of their PICC line (a peripherally inserted central catheter). A PICC is a line that is used to inject the patient with the chemical. This PICC line should be regularly cleaned, flushed to maintain patency and the insertion site checked for signs of infection. It takes approximately 10–15 minutes to take care of a PICC line by a nurse.
Cancer Care Manitoba provided access to the electronic scheduling system, also known as "ARIA" which is comprehensive information and image management system that aggregates patient data into a fully-­‐electronic medical chart, provided by VARIAN Medical System. This system was used to find out how many patients are booked in every clinic day. It also reveals which chair is used for how many hours. It was necessary to search a patient's history to find out how long the patient spends on which chair. Collecting the snapshot of each patient gives the complete picture of a one day clinic schedule.
The treatment room consists of the following two main limited resources:
Treatment Chairs: Chairs that are used to seat the patients during the treatment.
Nurses: Nurses are required to inject the treatment line into the patient and remove it at the end of the treatment. They also take care of the patients when they feel uncomfortable.
Mc Charles Chemotherapy unit consists of 11 nurses, and 5 stations with the following description:
Station 1: Station 1 has six chairs (numbered 1 to 6) and two nurses. The two nurses work from 8:00 to 16:00.
Station 2: Station 2 has six chairs (7 to 12) and three nurses. Two nurses work from 8:00 to 16:00 and one nurse works from 12:00 to 20:00.
Station 3: Station 4 has six chairs (13 to 18) and two nurses. The two nurses work from 8:00 to 16:00.
Station 4: Station 4 has six chairs (19 to 24) and three nurses. One nurse works from 8:00 to 16:00. Another nurse works from 10:00 to 18:00.
Solarium Station: Solarium Station has six chairs (Solarium Stretcher 1, Solarium Stretcher 2, Isolation, Isolation emergency, Fire Place 1, Fire Place 2). There is only one nurse assigned to this station that works from 12:00 to 20:00. The nurses from other stations can help when need arises.
There is one more nurse known as the "float nurse" who works from 11:00 to 19:00. This nurse can work at any station. Table 1 summarises the working hours of chairs and nurses. All treatment stations start at 8:00 and continue until the assigned nurse for that station completes her shift.
Currently, the clinic uses a scheduling template to assign the patients' appointments. But due to high demand of patient appointment it is not followed any more. We believe that this template can be improved based on the availability of nurses and chairs. Clinic workload was collected from 21 days of field observation. The current scheduling template has 10 types of appointment time slot: 15-­‐minute, 1-­‐hour, 1.5-­‐hour, 2-­‐hour, 3-­‐hour, 4-­‐hour, 5-­‐hour, 6-­‐hour, 8-­‐hour and 10-­‐hour and it is designed to serve 95 patients. But when the scheduling template was compared with the 21 days observations, it was found that the clinic is serving more patients than it is designed for. Therefore, the providers do not usually follow the scheduling template. Indeed they very often break the time slots to accommodate slots that do not exist in the template. Hence, we find that some of the stations are very busy (mostly station 2) and others are underused. If the scheduling template can be improved, it will be possible to bring more patients to the clinic and reduce their waiting time without adding more resources.
In order to build or develop a simulation model of the existing system, it is necessary to collect the following data:
Types of treatment durations.
Numbers of patients in each treatment type.
Arrival pattern of the patients.
Steps that the patients have to go through in their treatment journey and required time of each step.
Using the observations of 2,155 patients over 21 days of historical data, the types of treatment durations and the number of patients in each type were estimated. This data also assisted in determining the arrival rate and the frequency distribution of the patients. The patients were categorised into six types. The percentage of these types and their associated service times distributions are determined too.
ARENA Rockwell Simulation Software (v13) was used to build the simulation model. Entities of the model were tracked to verify that the patients move as intended. The model was run for 30 replications and statistical data was collected to validate the model. The total number of patients that go though the model was compared with the actual number of served patients during the 21 days of observations.
Improvement Scenarios
After verifying and validating the simulation model, different scenarios were designed and analysed to identify the best scenario that can handle more patients and reduces the average patient's waiting time. Based on the clinic observation and discussion with the healthcare providers, the following constraints have been stated:
The stations are filled up with treatment chairs. Therefore, it is literally impossible to fit any more chairs in the clinic. Moreover, the stakeholders are not interested in adding extra chairs.
The stakeholders and the caregivers are not interested in changing the layout of the treatment room.
Given these constraints the options that can be considered to design alternative scenarios are:
Changing the arrival pattern of the patients: that will fit over the nurses' availability.
Changing the nurses' schedule.
Adding one full time nurse at different starting times of the day.
Figure 2 compares the available number of nurses and the number of patients' arrival during different hours of a day. It can be noticed that there is a rapid growth in the arrival of patients (from 13 to 17) between 8:00 to 10:00 even though the clinic has the equal number of nurses during this time period. At 12:00 there is a sudden drop of patient arrival even though there are more available nurses. It is clear that there is an imbalance in the number of available nurses and the number of patient arrivals over different hours of the day. Consequently, balancing the demand (arrival rate of patients) and resources (available number of nurses) will reduce the patients' waiting time and increases the number of served patients. The alternative scenarios that satisfy the above three constraints are listed in Table 2. These scenarios respect the following rules:
Long treatments (between 4hr to 11hr) have to be scheduled early in the morning to avoid working overtime.
Patients of type 1 (15 minutes to 1hr treatment) are the most common. They can be fitted in at any time of the day because they take short treatment time. Hence, it is recommended to bring these patients in at the middle of the day when there are more nurses.
Nurses get tired at the end of the clinic day. Therefore, fewer patients should be scheduled at the late hours of the day.
In Scenario 1, the arrival pattern of the patient was changed so that it can fit with the nurse schedule. This arrival pattern is shown Table 3. Figure 3 shows the new patients' arrival pattern compared with the current arrival pattern. Similar patterns can be developed for the remaining scenarios too.
Analysis of Results
ARENA Rockwell Simulation software (v13) was used to develop the simulation model. There is no warm-­‐up period because the model simulates day-­‐to-­‐day scenarios. The patients of any day are supposed to be served in the same day. The model was run for 30 days (replications) and statistical data was collected to evaluate each scenario. Tables 4 and 5 show the detailed comparison of the system performance between the current scenario and Scenario 1. The results are quite interesting. The average throughput rate of the system has increased from 103 to 125 patients per day. The maximum throughput rate can reach 135 patients. Although the average waiting time has increased, the utilisation of the treatment station has increased by 15.6%. Similar analysis has been performed for the rest of the other scenarios. Due to the space limitation the detailed results are not given. However, Table 6 exhibits a summary of the results and comparison between the different scenarios. Scenario 1 was able to significantly increase the throughput of the system (by 21%) while it still results in an acceptable low average waiting time (13.4 minutes). In addition, it is worth noting that adding a nurse (Scenarios 3, 4, and 5) does not significantly reduce the average wait time or increase the system's throughput. The reason behind this is that when all the chairs are busy, the nurses have to wait until some patients finish the treatment. As a consequence, the other patients have to wait for the commencement of their treatment too. Therefore, hiring a nurse, without adding more chairs, will not reduce the waiting time or increase the throughput of the system. In this case, the only way to increase the throughput of the system is by adjusting the arrival pattern of patients over the nurses' schedule.
Developing a Scheduling Template based on Scenario 1
Scenario 1 provides the best performance. However a scheduling template is necessary for the care provider to book the patients. Therefore, a brief description is provided below on how scheduling the template is developed based on this scenario.
Table 3 gives the number of patients that arrive hourly, following Scenario 1. The distribution of each type of patient is shown in Table 7. This distribution is based on the percentage of each type of patient from the collected data. For example, in between 8:00-­‐9:00, 12 patients will come where 54.85% are of Type 1, 34.55% are of Type 2, 15.163% are of Type 3, 4.32% are of Type 4, 2.58% are of Type 5 and the rest are of Type 6. It is worth noting that, we assume that the patients of each type arrive as a group at the beginning of the hourly time slot. For example, all of the six patients of Type 1 from 8:00 to 9:00 time slot arrive at 8:00.
The numbers of patients from each type is distributed in such a way that it respects all the constraints described in Section 1.3. Most of the patients of the clinic are from type 1, 2 and 3 and they take less amount of treatment time compared with the patients of other types. Therefore, they are distributed all over the day. Patients of type 4, 5 and 6 take a longer treatment time. Hence, they are scheduled at the beginning of the day to avoid overtime. Because patients of type 4, 5 and 6 come at the beginning of the day, most of type 1 and 2 patients come at mid-­‐day (12:00 to 16:00). Another reason to make the treatment room more crowded in between 12:00 to 16:00 is because the clinic has the maximum number of nurses during this time period. Nurses become tired at the end of the clinic which is a reason not to schedule any patient after 19:00.
Based on the patient arrival schedule and nurse availability a scheduling template is built and shown in Figure 4. In order to build the template, if a nurse is available and there are patients waiting for service, a priority list of these patients will be developed. They are prioritised in a descending order based on their estimated slack time and secondarily based on the shortest service time. The secondary rule is used to break the tie if two patients have the same slack. The slack time is calculated using the following equation:
Slack time = Due time - (Arrival time + Treatment time)
Due time is the clinic closing time. To explain how the process works, assume at hour 8:00 (in between 8:00 to 8:15) two patients in station 1 (one 8-­‐hour and one 15-­‐ minute patient), two patients in station 2 (two 12-­‐hour patients), two patients in station 3 (one 2-­‐hour and one 15-­‐ minute patient) and one patient in station 4 (one 3-­‐hour patient) in total seven patients are scheduled. According to Figure 2, there are seven nurses who are available at 8:00 and it takes 15 minutes to set-­‐up a patient. Therefore, it is not possible to schedule more than seven patients in between 8:00 to 8:15 and the current scheduling is also serving seven patients by this time. The rest of the template can be justified similarly.
PMCID: PMC3562880  PMID: 23386870

Results 1-25 (2114094)