Automated Segmentation and Object Classification of CT Images: Application to In Vivo Molecular Imaging of Avian Embryos
Figure 2
Steps of Egg Detection and Shell Segmentation. (a) input image; (b) initial segmentation after coarse Gaussian blur; (c) slicewise inspection and search for round image objects; (c1) example of an image object (green) below the defined threshold for roundness; (c2) example of image objects above (white) and below (green) the defined threshold for roundness; (d) parts of the animal bed (orange) were removed by retaining only the round (white) image objects from the previous step and reexpanding them by applying a stringent surface tension criterion to prevent expansion too far back into the animal be; (e) the animal bed was separated from the background and expanded along the z-axis through the Coarse Egg/Bed (Parts) (white) object; (f) the Coarse Egg (white) object was segmented using a fixed threshold and further smoothed by three consecutive expansion and reduction operations; (g) using an automatically calculated threshold, the refined Egg object was segmented into Shell (light blue) and Interior (light red). The Shell object was surrounded by two additional layers of pixels (Shell Border, red).
