By Topic

Medical image retrieval: Multiple regression models for user's search target

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$31 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

2 Author(s)
Yue Li ; Coll. of Software, Nankai Univ., Tianjin, China ; Chia-Hung Wei

Breast cancer has been one of leading cancers in women around the world. A great number of digital mammograms are generated in hospitals and screening centers. Those digital mammograms can further be used for study and research by medical professionals. Content-based image retrieval refers to the retrieval of images whose contents are similar to a query example, using information derived from the images themselves. Relevance feedback, expressing the user's search target, can be used to bridge the semantic gap and improve the performance of CBIR systems. This study proposes a learning method for relevance feedback learning, which develops multiple logistic regression models to generalize the classification problem and provide an estimate of probability of class membership. To build the model, relevance feedback is utilized as the training data and the IRLS method is applied to estimate the parameters of the regression model and compute the maximum likelihood. Logistic regression models are created individually. After logistic regression models are fitted, discriminating features are selected by the measure of goodness of fit statistics. The weights of those discriminating features can be assigned according to their individual contributions to the maximum likelihood. The probability of the membership of the relevant class can therefore be obtained for each image of the database. Experimental results show that the proposed learning method can effectively improve the average precision from 30% to 65% through five iterations of relevance feedback rounds.

Published in:

Machine Learning and Cybernetics (ICMLC), 2011 International Conference on  (Volume:4 )

Date of Conference:

10-13 July 2011