Information Technology Reference
In-Depth Information
7
Mining Statistical Association Rules to Select
the Most Relevant Medical Image Features
Marcela X. Ribeiro
1
,AndreG.R.Balan
1
, Joaquim C. Felipe
2
,
Agma J.M. Traina
1
, and Caetano Traina Jr.
1
1
Department of Computer Science,
University of Sao Paulo at Sao Carlos, Brazil
2
Department of Physics and Mathematics,
University of Sao Paulo, at Ribeirao Preto, Brazil
Abstract.
In this chapter we discuss how to take advantage of association rule min-
ing to promote feature selection from low-level image features. Feature selection can
significantly improve the precision of content-based queries in image databases by re-
moving noisy and redundant features. A new algorithm named StARMiner is presented.
StARMiner aims at finding association rules relating low-level image features to high-
level knowledge about the images. Such rules are employed to select the most relevant
features. We present a case study in order to highlight how the proposed algorithm
performs in different situations, regarding its ability to select the most relevant fea-
tures that properly distinguish the images. We compare the StARMiner algorithm with
other well-known feature selection algorithms, showing that StARMiner reaches higher
precision rates. The results obtained corroborate the assumption that association rule
mining can effectively support dimensionality reduction in image databases.
7.1
Introduction
Nowadays, computational applications often need to deal with complex data,
such as images, video, time series, fingerprints, and DNA sequences. Focusing on
one of the most studied type of complex data - images, and more particularly on
medical images - two kinds of systems are now widely used: the Picture Archiving
and Communication Systems (PACS) and the Computer-Aided Diagnosis (CAD)
systems.
The CAD research community has been seeking, for several years, ecient and
precise algorithms to correctly classify medical images into relevant categories,
aiming at supporting medical diagnosis. The development of PACS broadens the
effective use of images on diagnosing, as well as in medicine teaching. However,
in order to be effectively useful, the processing of image retrieval in PACS and
CAD systems must also be fast and consistent with the judgment of specialists.
The volume of images generated on medical exams grows exponentially, de-
manding ecient and effective methods of image retrieval and analysis. In fact,
the number of images generated in hospitals and medical centers corresponds
to several Terabytes per day in a medium size hospital, what demands ecient
