Computer-Aided Diagnosis of Mammographic Calcification Clusters: Impact of Segmentation

Abstract

Medical image analysis is an area that has always attracted the interest of engineers and basic scientists. Research in the field has been intensified in the last 15-20 years. Significant work has been done and reported for breast cancer imaging with particular emphasis on mammography. The reasons for the impressive volume of work in this field include (a) increased awareness and education of women on the issues of early breast cancer detection and mammography, (b) the potential for significant improvements both in the fields of imaging and management, and (c) the multidisciplinary aspects of the problems and the challenge presented to both engineers and basic scientists.The importance of mammography and computer applications in mammography has been and continues to be the topic of numerous workshops, conferences, and publications [1,2]. There seems to be sufficient evidence that mammogra- 708Kallergi, Heine, and Tembey arguments that support an opposite view [3]. It should be noted that breast cancer was the second major cause of death for women in 2003 and mammography has been responsibly for a mortality reduction of 20-40% [4,5]. Despite its success, mammography still has a false negative rate of 10-30% and great variability [6].Calcifications are one of the main and earliest indicators of cancer in mammograms. They are present in 50-80% of all mammographically detected cancers but pathologic examinations reveal an even greater percentage [7]. Most of the minimal cancers and in-situ carcinomas are detected by the presence of calcifications [7]. A review of the literature on missed breast cancers indicates that calcifications are not commonly found among the missed lesions [8]. Although perception errors are not excluded, particularly in the case of microcalcifications (size < 1 mm), the technique of screen/film mammography (SFM) has been significantly improved over the years offering high-contrast and high-resolution mammograms that make calcification perception relatively easy. A greater and continuing problem for radiologists, with a major impact on the specificity of the diagnosis, is the mammographic differentiation between benign and malignant clustered calcifications. Almost all cases with calcifications are recommended for biopsy but only about 15-34% of these prove to be malignant [9]. The biopsies necessary to make the determination between benign and malignant disease represent the largest category of induced costs of mammography screening and a major source of concern for radiologists, surgeons, and patients. The advent of full field direct digital mammography will probably amplify this problem by providing more details and revealing breast abnormalities at very early stages [10].In the last 20 years, researchers have developed various computer schemes for analyzing mammograms with calcifications, masses, and other breast abnormalities in an effort to improve mammography and breast cancer detection and diagnosis [11]. Computer algorithms can be divided in three groups depending on their final goal:...

Image Segmentation for Evaluating Axillary Lymph Nodes

“…In addition, image segmentation was performed using symmlet wavelet and a new technique called the "donut filter." 25 ROIs were placed over the 138 benign calcifications and 122 microcalcifications, and the signals that emanated from these particular pixels were used to cue segmentation. The study provided promising results and should be replicated with images from a variety of digitizers as well as digital mammography.…”