In the past years, a great deal of effort was put into research regarding Indirect Immunofluorescence techniques with the aim of development of CAD systems. In this work a method for segmenting HEp-2 cells in IIF images is presented. Such task is one of the most challenging of automated IIF analysis, because the segmentation algorithm has to cope with a large heterogeneity of shapes and textures. In order to address this problem, numerous techniques and their combinations were evaluated, in a process aimed at maximizing the figure of merit. The proposed method, for a greater definition of cellular contours, uses the active contours in the last phase of the process. The initial conditions, center position and initial curve of the active contour, were obtained using a randomized Hough transform for ellipse; the idea in identifying cells was to approximate them initially to ellipses. The purpose of the active contours, within the segmentation process, is to allow the separation of connected regions (such as two overlapping cells), in order to obtain a better definition of the objects to be analyzed (the cells). Our system has been developed and tested on public database. Segmentation performances were evaluated in terms of Dice index and the method was compared with other state-of-the-art workers. The results obtained demonstrate the goodness of the method in the characterization of HEp-2 cells. The developed method shows great strength and convergence speed. Furthermore, the flexibility of the proposed method allows it to be easily used in other biomedical contexts.
|Title of host publication||ACM International Conference Proceeding Series|
|Number of pages||5|
|Publication status||Published - 2018|
All Science Journal Classification (ASJC) codes
- Human-Computer Interaction
- Computer Vision and Pattern Recognition
- Computer Networks and Communications