Title: A kernel approach for ensemble decision combinations with two-view mammography applications

Authors: Walker H. Land Jr., Dan Margolis, Maria Kallergi, John J. Heine

Addresses: Department of Bioengineering, Thomas J. Watson School of Engineering and Applied Science, Binghamton University, P.O. Box 6000, Binghamton, NY 13902-6000, USA. ' Department of Bioengineering, Thomas J. Watson School of Engineering and Applied Science, Binghamton University, P.O. Box 6000, Binghamton, NY 13902-6000, USA. ' Department of Medical Instruments Technology, Technological Educational Institute (TEI) of Athens, School of Technological Applications (STEF), Ag. Spyridonos Street, Egaleo 122 10 Athens, Greece. ' Cancer Prevention and Control Division, H. Lee Moffitt Cancer Center and Research Institute, 12902 Magnolia Dr., Tampa, FL 33612, USA

Abstract: An ensemble decision combination method was derived with kernel methods. An ensemble comprised of six Artificial Neural Networks (ANNs) was used to make benign–malignant predictions for breast lesions. The combination processing was evaluated for both single-view and two-view mammograms. The single-view combination showed marked improvements over the performance of each individual ANN, and the two-view combination showed marked improvements over the single-view combination performance. The work provides preliminary validation of the ensemble combination mechanism using an important clinically relevant data set.

Keywords: statistical estimation theory; statistical learning; ensemble combinations; experts mixture; automated mammographic diagnosis; classification; personalised medicine; differential evolution; kernel density; mammography; mammograms; benign lesions; artificial neural networks; ANNs; malignant lesions; breast lesions; cancer care strategies; breast cancer.

DOI: 10.1504/IJFIPM.2010.037152

International Journal of Functional Informatics and Personalised Medicine, 2010 Vol.3 No.2, pp.157 - 182

Received: 27 Apr 2010
Accepted: 16 Sep 2010

Published online: 29 Nov 2010 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article