Segmentation

Areas with nuclei in the image are detected. Active contours are used to segment each nucleus more precisely. 1. Original image 2. Detection of cells 3. Objects with an undefined edge are removed 4.Original image with "seeds" (masks) 5. Small objects are removed 6. Result: Automatic Segmentation 7. Result: Manual Segmentation

Automatic segmentation of nuclei in situ is one of the most technically challenging steps in digital image analysis. Segmentation of nuclei in monolayers prepared from suspensions is easy; a simple gray level thresholding will often do the trick. However, this monolayer technique requires laboratory preparations that physically isolate nuclei from their cells and tissue. Hence, important biological information is lost and the nuclear morphology is also affected. Alternatively, a successful digital segmentation will retain the biological information, save a lot of laboratory work and allow for cell-by-cell sequential multianalysis, among other analyses.

The most recent method for automatic segmentation involves several steps and combines multiple methods traditionally used. First, a local adaptive thresholding is used. To finely tune the segmentation we use active contours.