Authors:
Carolina Y. V. Watanabe
1
;
Marcela X. Ribeiro
2
;
Caetano Traina Jr.
3
and
Agma J. M. Traina
3
Affiliations:
1
Federal University of Rondônia, Brazil
;
2
Federal University of São Carlos, Brazil
;
3
University of São Paulo, Brazil
Keyword(s):
Statistical association rules, Computer-aided diagnosis, Associative classifier, Breast cancer.
Related
Ontology
Subjects/Areas/Topics:
Artificial Intelligence
;
Artificial Intelligence and Decision Support Systems
;
Biomedical Engineering
;
Business Analytics
;
Data Engineering
;
Data Mining
;
Databases and Information Systems Integration
;
Datamining
;
Enterprise Information Systems
;
Group Decision Support Systems
;
Health Information Systems
;
Multimedia Database Applications
;
Sensor Networks
;
Signal Processing
;
Soft Computing
Abstract:
In this paper, we present a new method called SACMiner for mammogram classification using statistical association rules. The method employs two new algorithms the StARMiner* and the Voting classifier (V-classifier). StARMiner* mines association rules over continuous feature values, avoiding introducing bottleneck and inconsistencies in the learning model due to a discretization step. The V-classifier decides which class best represents a test image, based on the statistical association rules mined. The experiments comparing SACMiner
with other traditional classifiers in detecting breast cancer in mammograms show that the proposed method reaches higher values of accuracy, sensibility and specificity. The results indicate that SACMiner is well-suited to classify mammograms. Moreover, the proposed method has a low computation cost, being linear on the number of dataset items, when compared with other classifiers. Furthermore, SACMiner is extensible to work with other types of medical i
mages.
(More)