skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: A Novel Local Learning based Approach With Application to Breast Cancer Diagnosis

Abstract

The purpose of this study is to develop and evaluate a novel local learning-based approach for computer-assisted diagnosis of breast cancer. Our new local learning based algorithm using the linear logistic regression method as its base learner is described. Overall, our algorithm will perform its stochastic searching process until the total allowed computing time is used up by our random walk process in identifying the most suitable population subdivision scheme and their corresponding individual base learners. The proposed local learning-based approach was applied for the prediction of breast cancer given 11 mammographic and clinical findings reported by physicians using the BI-RADS lexicon. Our database consisted of 850 patients with biopsy confirmed diagnosis (290 malignant and 560 benign). We also compared the performance of our method with a collection of publicly available state-of-the-art machine learning methods. Predictive performance for all classifiers was evaluated using 10-fold cross validation and Receiver Operating Characteristics (ROC) analysis. Figure 1 reports the performance of 54 machine learning methods implemented in the machine learning toolkit Weka (version 3.0). We introduced a novel local learning-based classifier and compared it with an extensive list of other classifiers for the problem of breast cancer diagnosis. Our experiments show that themore » algorithm superior prediction performance outperforming a wide range of other well established machine learning techniques. Our conclusion complements the existing understanding in the machine learning field that local learning may capture complicated, non-linear relationships exhibited by real-world datasets.« less

Authors:
 [1];  [1]
  1. ORNL
Publication Date:
Research Org.:
Oak Ridge National Lab. (ORNL), Oak Ridge, TN (United States)
Sponsoring Org.:
USDOE Laboratory Directed Research and Development (LDRD) Program
OSTI Identifier:
1036178
DOE Contract Number:  
DE-AC05-00OR22725
Resource Type:
Conference
Resource Relation:
Conference: SPIE Medical Imaging, San Diego, CA, USA, 20120204, 20120204
Country of Publication:
United States
Language:
English
Subject:
59 BASIC BIOLOGICAL SCIENCES; 99 GENERAL AND MISCELLANEOUS//MATHEMATICS, COMPUTING, AND INFORMATION SCIENCE; ALGORITHMS; BIOPSY; COMPUTER CALCULATIONS; DIAGNOSIS; FORECASTING; LEARNING; MAMMARY GLANDS; NEOPLASMS; PATIENTS; PERFORMANCE; VALIDATION

Citation Formats

Xu, Songhua, and Tourassi, Georgia. A Novel Local Learning based Approach With Application to Breast Cancer Diagnosis. United States: N. p., 2012. Web.
Xu, Songhua, & Tourassi, Georgia. A Novel Local Learning based Approach With Application to Breast Cancer Diagnosis. United States.
Xu, Songhua, and Tourassi, Georgia. Sun . "A Novel Local Learning based Approach With Application to Breast Cancer Diagnosis". United States.
@article{osti_1036178,
title = {A Novel Local Learning based Approach With Application to Breast Cancer Diagnosis},
author = {Xu, Songhua and Tourassi, Georgia},
abstractNote = {The purpose of this study is to develop and evaluate a novel local learning-based approach for computer-assisted diagnosis of breast cancer. Our new local learning based algorithm using the linear logistic regression method as its base learner is described. Overall, our algorithm will perform its stochastic searching process until the total allowed computing time is used up by our random walk process in identifying the most suitable population subdivision scheme and their corresponding individual base learners. The proposed local learning-based approach was applied for the prediction of breast cancer given 11 mammographic and clinical findings reported by physicians using the BI-RADS lexicon. Our database consisted of 850 patients with biopsy confirmed diagnosis (290 malignant and 560 benign). We also compared the performance of our method with a collection of publicly available state-of-the-art machine learning methods. Predictive performance for all classifiers was evaluated using 10-fold cross validation and Receiver Operating Characteristics (ROC) analysis. Figure 1 reports the performance of 54 machine learning methods implemented in the machine learning toolkit Weka (version 3.0). We introduced a novel local learning-based classifier and compared it with an extensive list of other classifiers for the problem of breast cancer diagnosis. Our experiments show that the algorithm superior prediction performance outperforming a wide range of other well established machine learning techniques. Our conclusion complements the existing understanding in the machine learning field that local learning may capture complicated, non-linear relationships exhibited by real-world datasets.},
doi = {},
journal = {},
number = ,
volume = ,
place = {United States},
year = {2012},
month = {1}
}

Conference:
Other availability
Please see Document Availability for additional information on obtaining the full-text document. Library patrons may search WorldCat to identify libraries that hold this conference proceeding.

Save / Share: