1Assistant Professor, Department of ECE, ANUCET, ANU, Guntur, A.P, India
2Professor, Department of ECE, Bapatla Engineering College, Bapatla, Guntur, A.P, India
*Corresponding Author: G. Prathibha, Assistant Professor, Dept. of ECE, ANUCET, Acharya Nagarjuna University. Email Id: prathibhamails@gmail.comwhich
Online published on 7 December, 2018.
An algorithm for classification of benign and malignant masses in digital mammograms is proposed in this paper. Feature vector is formulated based on the Diffusion Wavelets. Diffusion Wavelets are fast multi-scale framework for the analysis of functions on discrete (or discretize continuous) structures. Diffusion wavelets construct a compressed form of representation of the dyadic powers of a symmetric or non-symmetric square matrix by representing the associated matrices at each scale. Diffusion Wavelet coefficients are calculated for ROI's of preprocessed mammograms obtained from DDSM data base (Digital Database for Screening Mammography). Statistical parameters are calculated from Diffusion Wavelet Coefficients. The area under the curve Az=0.92 is achieved using KNN classifier for classification of malignant and benign ROI's of mammograms.
Mammograms, Diffusion Wavelets, KNN classifier, Area Under the Curve(AUC)
