IMAGE ANALYSIS OF HISTOPATHOLOGICAL IMAGES-USING AUTOMATIC SEGMENTATION OF CELL NUCLEI

                  IMAGE ANALYSIS OF HISTOPATHOLOGICAL   IMAGES-USING AUTOMATIC SEGMENTATION OF CELL NUCLEI

      

         ABSTRACT :

                         Automatic  segmentation  of  cell  nuclei  is  an  essential  step  in  image cytometry  and  histometry  .The  goal  of  this  study  is  to  develop  efficient  and accurate  algorithms for detecting and segmenting cell nuclei in 2-D histological images.  This  is  commonly  a  first  step  to  counting  cells,  quantifying  molecular markers  interest  in  healthy and pathologic  specimens  and also  for  quantifying aspects of normal/diseased tissue architecture. From the image,foreground  pixels  are  separated from the background pixels using a graph-cuts-based binarization. The  most  critical  aspect  of  nuclear  segmentation  algorithms  is  the  process  of detecting a set of points in the image,usually one per cell nucleus and close to its center, that are variously referred to as “markers” or “seeds.” The accuracy of the segmentation depends critically on the accuracy and reliability of the initial seed points. The initial segmentation is performed and it is refined by using a method of  alpha  expansions  and  graph  coloring. The  present    work  has  built  upon integrated,  and  extended  multiple  recent  advances  in  the  biological  image analysis field. The accuracy of the algorithm is investigated for the images with segmentation errors.