OpenCFU, a new free and open-source software to count cell colonies and other circular objects - PubMed (original) (raw)

OpenCFU, a new free and open-source software to count cell colonies and other circular objects

Quentin Geissmann. PLoS One. 2013.

Abstract

Counting circular objects such as cell colonies is an important source of information for biologists. Although this task is often time-consuming and subjective, it is still predominantly performed manually. The aim of the present work is to provide a new tool to enumerate circular objects from digital pictures and video streams. Here, I demonstrate that the created program, OpenCFU, is very robust, accurate and fast. In addition, it provides control over the processing parameters and is implemented in an intuitive and modern interface. OpenCFU is a cross-platform and open-source software freely available at http://opencfu.sourceforge.net.

PubMed Disclaimer

Conflict of interest statement

Competing Interests: The author has declared that no competing interests exist.

Figures

Figure 1

Figure 1. Flowchart representing the processing steps.

The image is preprocessed (1) in order to correct for gradual changes in background intensity and increase the contrast. The first pass of the processing (2) generates a score-map by iteratively annotating valid regions. The second pass (3) involves finding connected components in the thresholded score-map and segmenting them using a distance transform/watershed approach. Optional postprocessing filters (4) can be performed by OpenCFU or, using the raw data, by the user.

Figure 2

Figure 2. Illustration of the processing steps performed on three sample images.

Each channel of the original image (A) is preprocessed individually and merged to a grey-scale image (B). A score-map is generated by recursive thresholding an annotation of circular regions (C). This excludes regions that were morphologically unlikely to be colonies (i.e. arrows 1 and 2). The score-map is then thresholded by a user-defined or automatically calculated value (D). The objects identified as merged colonies (on the basis of their morphological features) are segmented using a watershed variant on their distance transform (E). Arrow 3 shows objects that have been successfully segmented. Finally, the morphologically valid objects can be assessed further using intensity and colour filters. Arrow 4 shows a minority contaminant bacteria that was excluded using postprocessing filter and represented by crossed-out red ellipses. Arrow 5 shows valid colonies represented by yellow and blue ellipses. For the purpose of explanation, only representative areas (200formula image200 pixels) of three processed images are shown here.

Figure 3

Figure 3. Processing time of OpenCFU, NICE and an ImageJ macro for images of different size.

An original arbitrary square image was either tiled to itself or scaled-up (A) in order to obtain a range of square images featuring an increasing number of colonies or increasingly large colonies, respectively. The processing time of the three methods for these images was estimated in both cases (B). On the tested range of resolutions, OpenCFU was faster than both NICE and the ImageJ macro (IJM). The segments joining points do not represent data, but only aid readability.

Figure 4

Figure 4. Comparison of accuracy between OpenCFU, NICE and an ImageJ macro .

The medians of seven humans counts were used as a reference to measure deviation. The effect of the number of colonies on the deviation from the reference was assessed (A). For NICE and the ImageJ macro (IJM), the slope was significantly negative. The dotted line represents the reference. The absolute deviation from the reference was used as a measure of error (B). Error for the best human, the worst human and the three methods were compared to the pooled human group. With high-definition images (HD), NICE and IJM had a higher error than the pooled human group (Pool) while OpenCFU (OFU) did not. Using low-definition pictures (LD) from a low-cost webcam increased the error for the three methods.

Figure 5

Figure 5. Comparison of robustness to common perturbations between OpenCFU, NICE and an ImageJ macro .

A qualitative assessment of robustness was undertaken by analysing pictures containing artefacts (A). Representative portions of 1.7 cm by 1.7 cm (200formula image200 pixels) illustrate the results of the presence of bubbles (1), cracks in the agar (2), dust (3) and edge of dish (4) in the region of interest. Objects detected by OpenCFU, NICE and the ImageJ macro (IJM) are represented by ellipses, crosses and arbitrary colours, respectively. NICE and IJM but not OpenCFU seemed to consistently detect artefacts as colonies. A quantitative analysis of robustness to plate mispositioning was conducted (B). OpenCFU, NICE and IJM were used to count the number of colonies in the pictures of 19 plates. Then, all the images were translated by 1.7 mm (25px) and analysed with the same region of interest as the original. This procedure induced a significant bias for NICE, formula image (formula image-formula image) colonies and IJM formula image (formula image-formula image) colonies, but not for OpenCFU formula image (formula image-formula image) colonies (one-sided paired t-test). The impact of the presence of bubbles in the agar was measured by analysing pictures of 18 plates containing exclusively bubbles (C). A linear regression between the number of bubbles and the number of detected objects was performed. NICE and IJM counts were both positively related to the number of bubbles, formula image (formula image-formula image) and formula image (formula image-formula image), respectively. OpenCFU was not affected: formula image (formula image-formula image.

Figure 6

Figure 6. Versatility of OpenCFU.

A qualitative assessment of the versatility of OpenCFU was undertaken by analysing pictures of different circular biological objects: a clear (A) and a poor quality (B) picture of Staphylococcus aureus colonies, a low-contrasted picture of Escherichia coli (C), a noisy picture of mustard seeds (D), a noisy picture of soy-bean seeds (E), and a micrography of Carduus sp. pollen (F). For the purpose of explanation, only representative areas (200formula image200 pixels) of six processed images are shown here. Original portions of images are on the left and correspond to the graphical results obtained using OpenCFU on the right.

Similar articles

Cited by

References

    1. Costa CM, Yang S (2009) Counting pollen grains using readily available, free image processing and analysis software. Annals of Botany 104: 1005–1010. - PMC - PubMed
    1. Mello CAB, dos Santos WP, Rodrigues MAB, Candeias ALB, Gusmo CMG (2008) Image segmen-tation of ovitraps for automatic counting of Aedes Aegypti eggs. Conference proceedings: Annual International Conference of the IEEE Engineering in Medicine and Biology Society IEEE Engi-neering in Medicine and Biology Society Conference 2008: 3103–3106. - PubMed
    1. Severini AD, Borrs L, Cirilo AG (2011) Counting maize kernels through digital image analysis. Crop Science 51: 2796.
    1. Forero MG, Pennack JA, Hidalgo A (2010) DeadEasy neurons: automatic counting of HB9 neuronal nuclei in Drosophila . Cytometry Part A: the journal of the International Society for Analytical Cytology 77: 371–378. - PubMed
    1. Kachouie NN, Kang L, Khademhosseini A (2009) Arraycount, an algorithm for automatic cell counting in microwell arrays. BioTechniques 47: x–xvi. - PMC - PubMed

Publication types

MeSH terms

Grants and funding

This study is part of the project “Multidrug resistance and the evolutionary ecology of insect immunity” (http://cordis.europa.eu/projects/rcn/96082_en.html) funded by the European Research Council. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

LinkOut - more resources