Hourglass Shapes in Rank Grey-Level Hit-or-miss Transform for Membrane Segmentation in HER2/neu Images

摘要

The paper presents an automatic approach to the analysis of images of breast cancer tissue stained with HER2 antibody. It applies the advanced morphological tools to build the system for recognition of the cell nuclei and the membrane localizations. The final results of image processing is the computerized method of estimation of the membrane staining continuity. The important point in this approach is application of the hourglass shapes in rank grey-level hit-or-miss transform of the image. The experimental results performed on 15 cases have shown high accuracy of the nuclei and membrane localizations. The mean absolute error of continuity estimation of the stained membrane between the expert and our system results was 6.1% at standard deviation of 3.2%. These results confirm high efficiency of the proposed solution.