An objective method to measure cell survival by computer-assisted image processing of numeric images of Petri dishes

A time-resolved clonogenic assay for improved cell survival and RBE measurements

Purpose

The in vitro clonogenic assay (IVCA) is the mainstay of quantitative radiobiology. Here, we investigate the benefit of a time-resolved IVCA version (trIVCA) to improve the quantification of clonogenic survival and relative biological effectiveness (RBE) by analyzing cell colony growth behavior.

Materials & Methods

In the IVCA, clonogenicity classification of cell colonies is performed based on a fixed colony size threshold after incubation. In contrast, using trIVCA, we acquire time-lapse microscopy images during incubation and track the growth of each colony using neural-net-based image segmentation. Attributes of the resulting growth curves are then used as predictors for a decision tree classifier to determine clonogenicity of each colony. The method was applied to three cell lines, each irradiated with 250 kV X-rays in the range 0-8 Gy and carbon ions of high LET (100 keV/μm, dose-averaged) in the range 0-2 Gy. We compared the cell survival curves determined by trIVCA to those from the classical IVCA across different size thresholds and incubation times. Further, we investigated the impact of the assaying method on RBE determination.

Results

Size distributions of abortive and clonogenic colonies overlap consistently, rendering perfect separation via size threshold unfeasible at any readout time. This effect is dose-dependent, systematically inflating the steepness and curvature of cell survival curves. Consequently, resulting cell survival estimates show variability between 3% and 105%. This uncertainty propagates into RBE calculation with variability between 8% and 25% at 2 Gy.Determining clonogenicity based on growth curves has an accuracy of 95% on average.

Conclusion

The IVCA suffers from substantial uncertainty caused by the overlap of size distributions of delayed abortive and clonogenic colonies. This impairs precise quantification of cell survival and RBE. By considering colony growth over time, our method improves assaying clonogenicity.

A framework for automated time-resolved analysis of cell colony growth after irradiation

Understanding dose-dependent survival of irradiated cells is a pivotal goal in radiotherapy and radiobiology. To this end, the clonogenic assay is the standard in vitro method, classifying colonies into either clonogenic or non-clonogenic based on a size threshold at a fixed time. Here we developed a methodological framework for the automated analysis of time course live-cell image data to examine in detail the growth dynamics of large numbers of colonies that occur during such an experiment. We developed a segmentation procedure that exploits the characteristic composition of phase-contrast images to identify individual colonies. Colony tracking allowed us to characterize colony growth dynamics as a function of dose by extracting essential information: (a) colony size distributions across time; (b) fractions of differential growth behavior; and (c) distributions of colony growth rates across all tested doses. We analyzed three data sets from two cell lines (H3122 and RENCA) and made consistent observations in line with already published results: (i) colony growth rates are normally distributed with a large variance; (ii) with increasing dose, the fraction of exponentially growing colonies decreases, whereas the fraction of delayed abortive colonies increases; as a novel finding, we observed that (iii) mean exponential growth rates decrease linearly with increasing dose across the tested range (0-10 Gy). The presented method is a powerful tool to examine live colony growth on a large scale and will help to deepen our understanding of the dynamic, stochastic processes underlying the radiation response in vitro.

Cell colony counter called CoCoNut

Clonogenic assays are powerful tools for testing cell reproductive death after biological damage caused by, for example, ionizing radiation. Traditionally, the methods require a cumbersome, slow and eye-straining manual counting of viable colonies under a microscope. To speed up the counting process and minimize those issues related to the subjective decisions of the scoring personnel, we developed a semi-automated, image-based cell colony counting setup, named CoCoNut (Colony Counter developed by the Nutech department at the Technical University of Denmark). It consists in an ImageJ macro and a photographic 3D-printed light-box, conceived and demonstrated to work together for Crystal Violet-stained colonies. Careful attention was given to the image acquisition process, which allows background removal (i.e. any unwanted element in the picture) in a minimally invasive manner. This is mainly achieved by optimal lighting conditions in the light-box and dividing the image of a flask that contains viable colonies by the picture of an empty flask. In this way, CoCoNut avoids using aggressive background removal filters that usually lead to suboptimal colony count recovery. The full method was tested with V79 and HeLa cell survival samples. Results were compared to other freely available tools. CoCoNut proved able to successfully distinguish between single and merged colonies and to identify colonies bordering on flask edges. CoCoNut software calibration is fast; it requires the adjustment of a single parameter that is the smallest colony area to be counted. The employment of a single parameter reduces the risk of subjectivity, providing a robust and user-friendly tool, whose results can be easily compared over time and among different bio-laboratories. The method is inexpensive and easy to obtain. Among its advantages, we highlight the possibility of combining the macro with a perfectly reproducible 3D-printed light-box. The CoCoNut software and the 3D-printer files are provided as supporting information (S1 CoCoNut Files).

Automated Wormscan

There has been a recent surge of interest in computer-aided rapid data acquisition to increase the potential throughput and reduce the labour costs of large scale Caenorhabditis elegans studies. We present Automated WormScan, a low-cost, high-throughput automated system using commercial photo scanners, which is extremely easy to implement and use, capable of scoring tens of thousands of organisms per hour with minimal operator input, and is scalable. The method does not rely on software training for image recognition, but uses the generation of difference images from sequential scans to identify moving objects. This approach results in robust identification of worms with little computational demand. We demonstrate the utility of the system by conducting toxicity, growth and fecundity assays, which demonstrate the consistency of our automated system, the quality of the data relative to manual scoring methods and congruity with previously published results.

WormScan: a technique for high-throughput phenotypic analysis of Caenorhabditis elegans

Background

There are four main phenotypes that are assessed in whole organism studies of Caenorhabditis elegans; mortality, movement, fecundity and size. Procedures have been developed that focus on the digital analysis of some, but not all of these phenotypes and may be limited by expense and limited throughput. We have developed WormScan, an automated image acquisition system that allows quantitative analysis of each of these four phenotypes on standard NGM plates seeded with E. coli. This system is very easy to implement and has the capacity to be used in high-throughput analysis.

Methodology/Principal Findings

Our system employs a readily available consumer grade flatbed scanner. The method uses light stimulus from the scanner rather than physical stimulus to induce movement. With two sequential scans it is possible to quantify the induced phototactic response. To demonstrate the utility of the method, we measured the phenotypic response of C. elegans to phosphine gas exposure. We found that stimulation of movement by the light of the scanner was equivalent to physical stimulation for the determination of mortality. WormScan also provided a quantitative assessment of health for the survivors. Habituation from light stimulation of continuous scans was similar to habituation caused by physical stimulus.

Conclusions/Significance

There are existing systems for the automated phenotypic data collection of C. elegans. The specific advantages of our method over existing systems are high-throughput assessment of a greater range of phenotypic endpoints including determination of mortality and quantification of the mobility of survivors. Our system is also inexpensive and very easy to implement. Even though we have focused on demonstrating the usefulness of WormScan in toxicology, it can be used in a wide range of additional C. elegans studies including lifespan determination, development, pathology and behavior. Moreover, we have even adapted the method to study other species of similar dimensions.

Automated counting of bacterial colony forming units on agar plates

Manual counting of bacterial colony forming units (CFUs) on agar plates is laborious and error-prone. We therefore implemented a colony counting system with a novel segmentation algorithm to discriminate bacterial colonies from blood and other agar plates.

A colony counter hardware was designed and a novel segmentation algorithm was written in MATLAB. In brief, pre-processing with Top-Hat-filtering to obtain a uniform background was followed by the segmentation step, during which the colony images were extracted from the blood agar and individual colonies were separated. A Bayes classifier was then applied to count the final number of bacterial colonies as some of the colonies could still be concatenated to form larger groups.

To assess accuracy and performance of the colony counter, we tested automated colony counting of different agar plates with known CFU numbers of S. pneumoniae, P. aeruginosa and M. catarrhalis and showed excellent performance.

Optimized digital counting colonies of clonogenic assays using ImageJ software and customized macros: comparison with manual counting

PURPOSE

To develop a digital method for counting colonies that highly replicates manual counting.

MATERIALS AND METHODS

Breast cancer cells were treated with trastuzumab-conjugated gold nanoparticles in combination with X-ray irradiation, (111)In labeled trastuzumab, or γ-radiation, followed by clonogenic assays. Colonies were counted manually or digitally using ImageJ software with customized macros. Key parameters, intensity threshold and minimum colony size, were optimized based on three preliminary manual counts or blindly chosen. The correlation of digital and manual counting and inter- and intra-experimenter variability were examined by linear regression. Survival curves derived from digital and manual counts were compared by F-test (P < 0.05).

RESULTS

Using optimized parameters, digital counts corresponded linearly to manual counts with slope (S) and R(2) value close to 1 and a small y-intercept (y(0)): SK-BR-3 (S = 0.96 ± 0.02, R(2) = 0.969, y(0) = 5.9 ± 2.2), MCF-7/HER2-18 (S = 0.98 ± 0.03, R(2) = 0.952, y(0) = 0.74 ± 0.47), and MDA-MB-231 cells (S = 1.00 ± 0.02, R(2) = 0.995, y(0) = 3.3 ± 4.5). Both reproducibility and repeatability of digital counts were better than the manual method. Survival curves generated from digital and manual counts were not significantly different; P-values were 0.3646 for SK-BR-3 cells and 0.1818 for MCF-7/HER2-18 cells. Using blind parameters, survival curves generated by both methods showed some differences: P-values were 0.0897 for SK-BR-3 cells and 0.0024 for MCF-7/HER2-18 cells.

CONCLUSIONS

The colony counting using ImageJ and customized macros with optimized parameters was a reliable method for quantifying the number of colonies.

Low-cost, high-throughput, automated counting of bacterial colonies

Research involving bacterial pathogens often requires enumeration of bacteria colonies. Here, we present a low-cost, high-throughput colony counting system consisting of colony counting software and a consumer-grade digital camera or document scanner. We demonstrate that this software, called "NICE" (NIST's Integrated Colony Enumerator), can count bacterial colonies as part of a high-throughput multiplexed opsonophagocytic killing assay used to characterize pneumococcal vaccine efficacy. The results obtained with NICE correlate well with the results obtained from manual counting, with a mean difference of less than 3%. NICE is also rapid; it can count colonies from multiple reaction wells within minutes and export the results to a spreadsheet for data processing. As this program is freely available from NIST, NICE should be helpful in bacteria colony enumeration required in many microbiological studies, and in standardizing colony counting methods.

Counting clonogenic assays from normoxic and anoxic irradiation experiments manually or by using densitometric software

The clonogenic assay is the method of choice to determine cell reproductive death after in vitro irradiation treatment. Traditionally, colony quantification has been performed by manual counting, a very laborious, time-consuming and rather subjective task. In this study, we compared manual counting by two skilled investigators with automated counting using the freely available ClonoCounter program. Five human tumour cell lines were irradiated under normoxia (21% O(2)) or anoxia (<0.1% O(2)), after 24 h or 6 h anoxic preincubation. Colonies were quantified manually or using the ClonoCounter software. A positive correlation between the absolute number of colonies counted manually and automatically was shown. Though there was a general trend of underpredicting the absolute number of cell colonies when counting automatically, survival curves were very similar, and in none of the cell lines were significant differences in radiobiological parameters such as mean inactivation dose, surviving fraction at 2 Gy and oxygen enhancement ratio detected. Our results suggest that the ClonoCounter provides sufficient reliability to be implemented for counting human tumour colonies in in vitro irradiation experiments. In contrast to several previously reported computer-aided colony-counting methods, it is a freely available program, requiring only minimal instrument costs.

Automated cell colony counting and analysis using the circular Hough image transform algorithm (CHiTA)

We present an automated cell colony counting method that is flexible, robust and capable of providing more in-depth clonogenic analysis than existing manual and automated approaches. The full form of the Hough transform without approximation has been implemented, for the first time. Improvements in computing speed have facilitated this approach. Colony identification was achieved by pre-processing the raw images of the colonies in situ in the flask, including images of the flask edges, by erosion, dilation and Gaussian smoothing processes. Colony edges were then identified by intensity gradient field discrimination. Our technique eliminates the need for specialized hardware for image capture and enables the use of a standard desktop scanner for distortion-free image acquisition. Additional parameters evaluated included regional colony counts, average colony area, nearest neighbour distances and radial distribution. This spatial and qualitative information extends the utility of the clonogenic assay, allowing analysis of spatially-variant cytotoxic effects. To test the automated system, two flask types and three cell lines with different morphology, cell size and plating density were examined. A novel Monte Carlo method of simulating cell colony images, as well as manual counting, were used to quantify algorithm accuracy. The method was able to identify colonies with unusual morphology, to successfully resolve merged colonies and to correctly count colonies adjacent to flask edges.

Counting colonies of clonogenic assays by using densitometric software

Clonogenic assays are a useful tool to test whether a given cancer therapy can reduce the clonogenic survival of tumour cells. A colony is defined as a cluster of at least 50 cells which can often only be determined microscopically. The process of counting colonies is very extensive work and so we developed software that is able to count the colonies automatically from scanned flasks. This software is made freely available by us with a detailed description how to use and install the necessary features.

Scoring CFU-GM colonies in vitro by data fusion: A first account

OBJECTIVE

In vitro models of hematopoiesis used in investigative hematopathology and in safety studies on candidate drugs, involve clonogenic assays on colony-forming unit granulocyte macrophage (CFU-GM). These assays require live and unstained colonies to be counted. Most laboratories still rely on visual scoring, which is time-consuming and error-prone. As a consequence, automated scoring is highly desired. An algorithm that recognizes and scores CFU-GM colonies by data fusion has been developed. Some preliminary results are presented in this article.

METHODS

CFU-GM assays were carried out on hematopoietic progenitors (human umbilical cord blood cells) grown in methylcellulose. Colony images were acquired by a digital camera and stored.

RESULTS

The classifier was designed to process images of layers sampled from a three-dimensional (3D) domain and forming a stack. Structure and texture information was extracted from each image. Classifier training was based on a 3D colony model applied to the image stack. The number of scored colonies (assigned class) was required to match the count supplied by the human expert (class of belonging). The trained classifier was validated on one more stack and then applied to a stack with overlapping colonies. Scoring in distortion- and caustic-affected border areas was also successfully demonstrated. Because of hardware limitations, compact colonies in some cases were missed.

CONCLUSIONS

The industry's scoring methods all rely on structure alone and process 2D data. Instead, the classifier here fuses data from a whole stack and is capable, in principle, of high-throughput screening.

Synchrotron radiation-based experimental determination of the optimal energy for cell radiotoxicity enhancement following photoelectric effect on stable iodinated compounds

This study was designed to experimentally evaluate the optimal X-ray energy for increasing the radiation energy absorbed in tumours loaded with iodinated compounds, using the photoelectric effect. SQ20B human cells were irradiated with synchrotron monochromatic beam tuned at 32.8, 33.5, 50 and 70 keV. Two cell treatments were compared to the control: cells suspended in 10 mg ml⁻¹ of iodine radiological contrast agent or cells pre-exposed with 10 μM of iodo-desoxyuridine (IUdR) for 48 h. Our radiobiological end point was clonogenic cell survival. Cells irradiated with both iodine compounds exhibited a radiation sensitisation enhancement. Moreover, it was energy dependent, with a maximum at 50 keV. At this energy, the sensitisation calculated at 10% survival was equal to 2.03 for cells suspended in iodinated contrast agent and 2.60 for IUdR. Cells pretreated with IUdR had higher sensitisation factors over the energy range than for those suspended in iodine contrast agent. Also, their survival curves presented no shoulder, suggesting complex lethal damages from Auger electrons. Our results confirm the existence of the 50 keV energy optimum for a binary therapeutic irradiation based on the presence of stable iodine in tumours and an external irradiation. Monochromatic synchrotron radiotherapy concept is hence proposed for increasing the differential effect between healthy and cancerous tissue irradiation.