Image processing methods implemented in AimSeg

AimSeg is implemented as a Fiji workflow and maintained through an update site. AimSeg handles the segmentation output as three different types of data, including binary masks, ROIs and label masks. Basic binary operations (erode, dilate, open, close, fill) are implemented as modifications of ImageJ 1.x source code [48]. Logical operators use ImageJ’s ImageCalculator class. Binary reconstruction is part of the morphological operations provided at Fiji’s Morphology update site [49]. Connected components from binary masks are detected, filtered and converted into ROIs with the ImageJ ParticleAnalyzer plugin. AimSeg handles different ROI sets by means of independent ImageJ RoiManager instances. The RoiManager is also used to transform ROIs into binary or label masks. Operations with individual ROIs use ImageJ Roi, PolygonRoi and ShapeRoi classes. These include calculating the convex hull, filling of ShapeRois, filtering PolygonRois contained in ShapeRois by specified criteria, and calculating the intersection of two ROIs. Additionally, ROI erosion and dilation is performed with ImageJ RoiEnlarger plugin. Operations with label masks are implemented using the MorphoLibJ library [50] by accessing the MarkerControlledWatershedTransform2D and ReplaceLabelValues classes. Label masks are transformed into ROIs by means of the ImageJ’s ThresholdToSelection plugin.