• 2018-07
  • 2018-10
  • 2018-11
  • 2019-04
  • 2019-05
  • 2019-06
  • 2019-07
  • 2019-08
  • 2019-09
  • 2019-10
  • 2019-11
  • 2019-12
  • 2020-01
  • 2020-02
  • 2020-03
  • 2020-04
  • 2020-05
  • 2020-06
  • 2020-07
  • CellProfiler and OpenComet are two open source software


    CellProfiler and OpenComet are two open source software available in internet to analyse silver stained comet assay images. With CellProfiler [19], Gonzalez et al. [18] modified the pipeline provided in the website by adding a background subtraction module and two filtering modules to quantify DNA damage. The performance of the CellProfiler software was evaluated with all the 40 cases under study and the TP (%) (True Positive) obtained was only 46.15%. Gyori et al. [12] developed OpenComet for the analysis of comet assay images. It is developed as a plug-in for the image processing platform, ImageJ. In OpenComet, comet segmentation is based on geometric shape attributes and comet partitioning is based on image intensity profile analysis. This tool can be used for the analysis of both fluorescent and silver stained images.
    Materials and methods
    Discussion An automated method for DNA damage analysis using silver stained comet assay images for clinical application is described in this paper. The silver stained comet assay images are very noisy when compared with fluorescent stained images and hence ordinary detection algorithms fail in effectively detecting the comets. Therefore, in the comet segmentation stage, shading correction has been incorporated with morphological bottom hat transformation. Two Niflumic acid synthesis enhancement stages have been included along with Gaussian filtering to enhance the comets against the background. Then by using thresholding and morphological operations, comets are identified from the silver stained comet assay images. In the comet partitioning stage, the individual comets are segmented into four regions as head, halo, tail and background using FCM. Then the output of FCM has been modified with the proposed clustering and partitioning algorithms. In the comet quantification stage, the DNA (%) in tail is considered for quantitatively analysing the DNA damage. The performance of the proposed method is compared with that of a recent method [12]. First and second rows of Fig. 5 show the results of comet segmentation stage of the proposed method and that of Gyori et al. [12], respectively. The comets selected by the expert are indicated with a yellow star along with those of the proposed method. OpenComet [12] is good at selecting the comets with minimum tail loss. Comet segmentation is based on shape attributes and hence, some noisy structures having similar shapes as that of comets are also detected as actual comets (refer yellow circles in Fig. 5). Therefore FP (%) is high using this method. There are only nine true comets present in Fig. 5(a). But 14 comets are detected by OpenComet. One of the true comets (comet indicated with green circle) has been rejected. Some of the true comets are selected as outliers (as indicated with rose circles in Figs. 5(d) and (f)) which reduced the TP (%). The five performance indices are tabulated in Table 1. Compared to OpenComet, a high improvement in PPV and in sensitivity is obtained. The slight outperforming of OpenComet in Case 2 for TP (%) is due to the area covered by each comet in the proposed method is slightly larger than that is covered by OpenComet and hence comets very near to the edges are rejected. This gives a lower TP (%) with the proposed method. An additional facility is provided in OpenComet to reject wrongly identified comets using manual selection. But, it cannot be tuned by setting any parameters. In comet partitioning phase, the algorithm should correctly detect head and tail regions. In the proposed method, FCM clustering along with proposed modifications results in better head and tail selection. Choosing the number of clusters is important. If we choose 3 clusters as head, tail and background, the DNA (%) in tail obtained is greater than that expected (in the case of good and moderately damaged cells). Therefore 4 clusters have been selected and in the modified clustering algorithm, final head region is identified as head and halo regions together. The results thus obtained are comparable to the values expected. In OpenComet, comet head is detected incorrectly in some cases, which is indicated in Figs. 5(b) and (f) with blue circles. Head detection based on profile analysis can fail in the case of silver stained images. In some cases head position is wrongly detected as indicated in Fig. 5(b) and in some other cases head area selected is larger than expected which is indicated in Fig. 5(f). The proposed method is tested for all types of cells having different levels of damage and in all cases, head is detected correctly (refer Fig. 6(d)).