I chose values of 30 for each, which greatly improved the segmentation results, as you can see in this comparison:Ĭomparison of the automatic settings for declumping (left) vs. In order to avoid this oversegmentation, I set the smoothing filter and the minimum distance between local maxima manually, an adjustment we often need to make for fibrillar structures (see the module help settings or our video on how exactly segmentation works in CellProfiler for more information). background, I noted that many of the longer fibrils were oversegmented (broken into too many pieces). While I was happy with the thresholding of fibril vs.
Since the image that we’re segmenting on has been enhanced and masked, I used the Workspace viewer to assess the segmentation while looking at the raw image to make sure my preprocessing steps hadn’t added or removed any “real” signal: Happily, the default settings worked well for thresholding fibers. See this blog post for an explanation of how the Robust Background method works. Since most of the image that we’re segmenting is background, I started by testing the Robust Background method for segmentation. Once I identify the fibrils as objects, I can then measure many different attributes such as area, length, width, etc. In this post, I’ll explain how I detected the fibrils as individual objects using IdentifyPrimaryObjects. If you’d like to follow along in CellProfiler, the pipeline and images for this project are available here. In combination with masking out very bright debris pixels ( Part II of this series), these steps helped prepare this image for segmentation of the fibrils. In the previous post in this “Thinking like an image analyst” series, I explained how I enhanced fibers to increase their brightness and applied a background subtraction to decrease the intensity of the background.
0 kommentar(er)
