Automatic Classification of Masses in Mammographic Images using Sparse Representation

Please enable javascript in your browser.

Zarghami, Ali | 2013

695 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 44701 (19)
University: Sharif University of Technology
Department: Computer Engineering
Advisor(s): Manzouri, Mohammad Taghi
Abstract:
Computer Aided Diagnosis (CAD) systems are widely used in different medical tasks. Radiology is a branch of medicine which takes advantage of image processing techniques to help radiologists, analyse complicated radiologic images. Among all kind of medical imagingprocedures, utilization of screening mammographyisgetting very popularin detection of breast abnormalities. A typical CAD system for mammogram analysis uses image enhancement and segmentation as pre-processing phase, and feature extraction and classification for detection phase. In this thesis, we have studied different approaches in each level of image processing required in a mammogram mass classification systems, and introduced a new approach to distinguish benign and malignant masses. A new feature of image texture is introduced and used along with a sparse classifier to classify the masses. Empirical results show that our approach has a success rate identical to the state of the art approaches with a great advantage of not using the exact boundary of the mass in the mammogram image which would be very difficult to detect in some instances otherwise
Keywords:
Radiology ; Breast Cancer ; Mammography ; Computer Aided Diagnosis Medicine ; Sparse Representation

No TOC