Classifying ductal trees using geometrical features and ensemble learning techniques
(2013) In Communications in Computer and Information Science 384. p.146-155- Abstract
Early detection of risk of breast cancer is of upmost importance for effective treatment. In the field of medical image analysis, automatic methods have been developed to discover features of ductal trees that are correlated with radiological findings regarding breast cancer. In this study, a data mining approach is proposed that captures a new set of geometrical properties of ductal trees. The extracted features are employed in an ensemble learning scheme in order to classify galactograms, medical images which visualize the tree structure of breast ducts. For classification, three variants of the AdaBoost algorithm are explored using as weak learner the CART decision tree. Although the new methodology does not improve the... (More)
Early detection of risk of breast cancer is of upmost importance for effective treatment. In the field of medical image analysis, automatic methods have been developed to discover features of ductal trees that are correlated with radiological findings regarding breast cancer. In this study, a data mining approach is proposed that captures a new set of geometrical properties of ductal trees. The extracted features are employed in an ensemble learning scheme in order to classify galactograms, medical images which visualize the tree structure of breast ducts. For classification, three variants of the AdaBoost algorithm are explored using as weak learner the CART decision tree. Although the new methodology does not improve the classification performance compared to state-of-the-art techniques, it offers useful information regarding the geometrical features that could be used as biomarkers providing insight to the relationship between ductal tree topology and pathology of human breast.
(Less)
- author
- Skoura, Angeliki ; Nuzhnaya, Tatyana ; Bakic, Predrag R. LU and Megalooikonomou, Vasilis
- publishing date
- 2013
- type
- Contribution to journal
- publication status
- published
- subject
- keywords
- Breast imaging, Classifier ensembles, Feature extraction
- in
- Communications in Computer and Information Science
- volume
- 384
- pages
- 10 pages
- publisher
- Springer
- external identifiers
-
- scopus:84885136531
- ISSN
- 1865-0929
- DOI
- 10.1007/978-3-642-41016-1_16
- language
- English
- LU publication?
- no
- id
- 5c48a4c3-3e07-4ba1-9e1f-9176ae9cd32a
- date added to LUP
- 2020-11-07 13:11:49
- date last changed
- 2022-02-01 17:37:39
@article{5c48a4c3-3e07-4ba1-9e1f-9176ae9cd32a,
abstract = {{<p>Early detection of risk of breast cancer is of upmost importance for effective treatment. In the field of medical image analysis, automatic methods have been developed to discover features of ductal trees that are correlated with radiological findings regarding breast cancer. In this study, a data mining approach is proposed that captures a new set of geometrical properties of ductal trees. The extracted features are employed in an ensemble learning scheme in order to classify galactograms, medical images which visualize the tree structure of breast ducts. For classification, three variants of the AdaBoost algorithm are explored using as weak learner the CART decision tree. Although the new methodology does not improve the classification performance compared to state-of-the-art techniques, it offers useful information regarding the geometrical features that could be used as biomarkers providing insight to the relationship between ductal tree topology and pathology of human breast.</p>}},
author = {{Skoura, Angeliki and Nuzhnaya, Tatyana and Bakic, Predrag R. and Megalooikonomou, Vasilis}},
issn = {{1865-0929}},
keywords = {{Breast imaging; Classifier ensembles; Feature extraction}},
language = {{eng}},
pages = {{146--155}},
publisher = {{Springer}},
series = {{Communications in Computer and Information Science}},
title = {{Classifying ductal trees using geometrical features and ensemble learning techniques}},
url = {{http://dx.doi.org/10.1007/978-3-642-41016-1_16}},
doi = {{10.1007/978-3-642-41016-1_16}},
volume = {{384}},
year = {{2013}},
}