A Novel Approach to Detect Microcalcification for Accurate Detection for Diagnosis of Breast Cancer

Abstract

Breast Cancer is one in all the top cancers inflicting death of assorted woman’s across the globe. It is really dangerous but at the same can be cured completely in case of early detection of it happens. Worldwide radiologist and doctors are working on various techniques and they found that mammographic identification of breast cancer images can be important in early detection of breast tissue cancer. Primary and main indication of the breast cancer is presenting the clusters of microcalcification. Although it becomes very difficult and also time consuming to make the arrangement of microcalcifications by the radiologists as benign or malignant. Also, the interpretation of mammograms isn’t straightforward thanks to tiny variations in densities of various tissues at intervals the image. It’s absolutely matching case for dense breast. This work presents novel approach that can perform classification of the clusters which is part of micro classification presents in the mammograms. In case of multiscale morphology, it observed the inter-connectivity of each microcalcification. This work is different one from the existing which mainly has attainment on the morphology of the microcalcification and the cluster feature for representing the cluster structure of microcalcification it generates a graphs. Theoretical features of a graph are extracted to maintain topological features used to make classification of the clusters of microcalcification using the k-nearest neighbour algorithm. This projected work evaluated with the assistance of 2 most accepted digitized dataset referred to as MIAS and DDSM and capable digital dataset. This work shows the different topology modeling is extremely necessary for accurate estimation of the microcalcification. Microcalcification thanks to the improved accuracy of microcalcification and conjointly the topological measures that is in a position to couple with clinical understanding.