Freeware tool for analysing numbers and sizes of cell colonies

Abstract

The clonogenic cell survival assay is a basic method to study the cytotoxic effect of radiation and chemical toxins. In large experimental setups, counting of colonies by eye is tiresome and prone to bias. Moreover, it is often interesting to quantify the size of individual colonies. Such analyses are largely facilitated by computerised image analysis systems. Although a number of such systems exist, they all focus on enumerating colonies and not on analysing the colony size. We have developed a new software package for both counting colonies and plotting their size distributions. The software called count and Plot HIstograms of Colony Size (countPHICS) consists of two parts: (1) a macro written for ImageJ which analyses computerised images of cell culture dishes or 6-well plates, counts colonies, estimates their size and saves the results in a text file; (2) a program written with QT Creator which reads the text file, plots histograms of colony size distribution and fits the best function. The full program is freely available at: http://www.fuw.edu.pl/~bbrzozow/FizMed/countPHICS.html . In conclusion, our new publically available software will facilitate colony counting and provide additional information on the colony growth rate, which is relevant especially for radiosensitisation studies.

Keywords: Clonogenic cell survival assay; Colony counter; Colony size histogram; countPHICS.

Fig. 1

Comparison of the pixel value distribution for different stack image components. White colour in RGB scale is (255, 255, 255) while black is (0, 0, 0). The first channel is red (R), the second—green (G) and the third—blue (B; RGB). To obtain the highest contrast, the channel where the average colony pixel value differed most from the background was selected. As shown in the RGB image, the colony is mostly blue in this case, which is also displayed in the graph. In this particular image, background was more white than black (closer to (255, 255, 255) than (0, 0, 0)), therefore the channel with the lowest average pixel value was chosen. In our case, it was green. The lower figure panels demonstrate that the green channel indeed gives far better contrast than the blue. In countPHICS, the standard deviation (σ) parameter was chosen for channel selection, because it distinguishes which channel differs the most from the background in general. (Colour figure online)

Fig. 2

Parameter sigma (standard deviation of the Gaussian blurring) and rolling ball radius (background subtraction) were optimised to get the best agreement between number of colonies counted manually and automatically in relation to image resolution (dots per inch, DPI)

Fig. 3

Watershed algorithm effect in the original picture (a) and the picture after Gaussian blurring (b). Green contours show the proper separation of the merged colonies which appear independently of Gaussian blurring application. Without Gaussian blurring, the watershed algorithm incorrectly divides a single colony into pieces (red contours). (Colour figure online)

Fig. 4

An example of a normalized colony size histogram with calculated Weibull parameters: mean value (μ) and square root of variation (σ)

Fig. 5

Representative images of the chosen categories: #1, #2, #3 and #4

Fig. 6

The size of colonies scored automatically (left) and manually (right) and compared with the Weibull fit for four different sample categories: #1, #2, #3 and #4. The number of analysed plates per sample category is given in Table 1

Fig. 7

Histograms using colony sizes for all 3 experiments (a) and the mean colony size (+/-standard deviation, 3 experiments) (b) for A549 tumour initiating cells (TICs): untreated, after MEK inhibition, gamma irradiation (2 Gy) and the combination of these two. A reduction in growth rate is observed