Digital Image Processing

Abstract

The paper discusses an approach involving digital image processing for estimating the breast cancer cell population in breast tissue sample. The process aims at providing a reliable,repeatable, and fast method that could replace the traditional methodof manual examination and subsequent estimation. The marker discussed in the paper is the Estrogen Receptor (ER) that gives a clear indication of the presence of cancer cells in the tissue sample. The methods involved are HSV color conversion from RGB image, Hue, saturation and value based object-background separation, morphological operations such as dilation and closing, and area based filtering for preliminary preparation of image for detailed analysis. A modified watershed algorithm designed for eliminating errors arising due to over-segmentation in traditional watershed algorithm is proposed to provide comparatively more accurate results. Further, intensity based thresholding is performed for identifying andcategorizing the cancerous cells into levels of severity of damage done to cells due to cancer. The proposed modified watershed algorithm is compared with the original watershed algorithm and an accuracy of almost 96.44% was observed and verified.

Keywords

Full Text