Using Reference Points for Detection of Calcifications in Mammograms for Medical Active Systems
Conference Paper
Overview
Research
Identity
Additional Document Info
View All
Overview
abstract
Context-awareness is one of the main features of any AmI system, particularly related to healthcare systems. To develop this kind of active systems it is necessary to create methodologies where it can be observed how the experts solve specific problems or situations. This paper focuses on the problem of breast cancer detection and classification, and presents a methodology based on a case study where radiologists and medical doctors were involved and their knowledge was extracted in order to define a model that could be used to develop a breast cancer automatic interpretation system, with the goal of helping the medical personnel in their decision-making process. As a first step towards the development of such active system, this paper presents the detection of calcifications, a type of finding in the breasts that could be benign or malign, depending on certain features. The detection is made by applying composite correlation filters, and even though the calcifications used for testing cover all the categories defined by a medical taxonomy (BI-RADS system), the accuracy of such detection is promising, where most categories have a detection accuracy of 80%25 or above.
