Automated Completion of Segmented Fragments of Macrophages using Weighted Dilation and Erosion in 2D+time Microscopy Videos

This paper presents a method for automated completion of segmented fragments in 2D+time microscopy movies. This method aims to automatically and accurately compute morphological features for segmented macrophages that are critical for understanding their phenotypic characteristics. 
In the situation that segmentation results in fragments corresponding to a single macrophage, it becomes challenging to determine whether these fragments belong to the same macrophage. Consequently, it is essential to complete the fragments for accurate quantitative analysis of morphology.
To achieve this, we propose a method based on weighted dilation and erosion (WDE) in the level-set formulation. By regulating the speed of the level lines based on local image intensity and thresholds calculated from the local Otsu's method, this approach effectively integrates fragmented segments while preserving the overall macrophage shape. The efficacy of the method is demonstrated through both visual and quantitative assessments, which indicate its ability to accurately complete segmented macrophages across a range of shapes and intensity levels in the images. Furthermore, the method has been shown to improve the accuracy of quantitative assessments when compared to ground truth images.