Novel segmentation algorithm for identification of cell membrane staining in HER2 images

The paper presents new approach to the problem of a membrane staining determination in HER2 images of breast cancer. It is based on representation of the small membrane sections by the defined linear patterns of different shapes, which combined together allow representing the complex shape of the membrane staining. The discovered linear structures existing in the image are localized and recognized applying the specially devised masks sliding over the whole analyzed image. The method is resistant to the noise inherent in the image and to the varying detail complexity of the image. Its usefulness in HER2 image processing was verified on examples of the HER2 images of the breast cancer representing different grades of membrane staining.