Novel algorithm for estimation of prognostic Ki-67 factor in selected view fields in breast cancer microscopic images

Following paper presents new approach to the problem of automotive Ki-67 proliferation factor estimation in breast cancer biopsy images. This method is based on context filtering with designated neighborhood masks for edge detection. Proposed method was designed to compensate two major problems: cell shape divergence from ellipse and variability of color hue along with intensity in cell staining. Both factors cause issues with the edge detection and segmentation. Proposed solution exhibits absolute mean error value equal to 6.5% in comparison to expert results.