Eventos Anais de eventos
ENEBI 2018
6º Encontro Nacional de Engenharia Biomecânica
COMPARAÇÃO DO USO DE VARIÁVEIS CONTÍNUAS E VARIÁVEIS INTERVALARES NA CLASSIFICAÇÃO DE PATOLOGIAS EM TERMOGRAMAS DE MAMA
Submission Author:
Marcus Costa de Araujo , PE
Co-Authors:
Marcus Costa de Araujo, Rita de Cássia Fernandes de Lima, Renata Souza
Presenter: Marcus Costa de Araujo
doi://10.26678/ABCM.ENEBI2018.EEB18-0200
Abstract
Thermography has became as a potential method to aid in the early detection of breast cancer. Recent studies involving the use of thermography as a screening technique have generated increasing interest, especially in cases where mammography is limited, as for young patients with dense breast tissue. The aim of this study is to evaluate the feasibility of using interval data from the perspective of the symbolic data analysis (SDA) to classify breast images as malignant, benignant or cyst abnormalities. Two approaches are compared for the task of classifying breast abnormalities from thermography. In the first one, continuous variables and a minimum distance classifier are used; the second approach uses a classifier based on interval data. The first approach obtained 18% of classification error rate, with 85% sensitivity for the malignant class. The second approach achieved a 20% error rate with 100% sensitivity to the malignant class.
Keywords
Termografia por Infravermelho, Detecção de câncer de mama, Classificação, Dados simbólicos
