Figure 1.
Process flow diagram for the proposed approach.
A preprocessing step is used to remove the background and uniform the light intensity conditions, then the segmentation process takes place with a two-dimensional version of the Enhanced Interaction Model. The resulting binary image is then enhanced through a cascade of morphological operators. Different colonies are then processed through a Watershed transform which returns as output the segmentation for each colony in the image.
Figure 2.
(a): original image, (b): background subtraction. (c): colony segmentation, (d, e): orientation matching. (f): output image. Segmentation, identification and other outputs are shown overimposed on the original image (for a description see Figure 3).
Table 1.
Table of the parameters adopted for the analysis.
Figure 3.
(a, b, c): original images. (d, e, f): segmented colonies and identified spotted cells. Colonies are segmented with different colors: in each one the central blue star point is the centroid while small green circles identify marked cells; the segment from the centroid to the marked cell defines the normalization path, whose value is shown next to the small circle.
Figure 4.
Distribution for the statistical analysis on the Zscan4-marked ES colony image set.
Distribution related to the distances from the centroid compared to the null case distribution, with a statistical difference p-value of with significance level
.
Figure 5.
Diagrams for the statistical analysis on the Zscan4-marked cells mutual distances.
Distribution related to the mutual distances compared to the null case distribution, with a p-value of with significance level
.
Figure 6.
Union processing of two channels for the fluorescence microscopy images (this is from the 40 nanog-marked colony image set). (a): DAPI channel (blue), (b): Nanog channel (green), (c): resulting image; it was preprocessed so the colors were similar to those in the Zscan4 set so that the algorithm could be applied immediately.
Figure 7.
Diagrams for the statistical analysis on the EndoA-marked ES colony image set.
Distribution related to the distances from the centroid compared to the null case distribution, with a statistical difference p-value of with significance level
.
Figure 8.
Diagrams for the statistical analysis on the EndoA-marked cell mutual distances.
Distribution related to the mutual distances compared to the null case distribution, with a p-value of with significance level
.
Figure 9.
Diagrams for the statistical analysis on the Nanog-marked ES colony image set.
Distribution related to the distances from the centroid compared to the null case distribution, with a statistical difference p-value of with significance level
.
Figure 10.
Diagrams for the statistical analysis on the Pou5f1-marked ES colony image set.
Distribution related to the distances from the centroid compared to the null case distribution, with a statistical difference p-value of with significance level
.