HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text Full Text (PDF) Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Chen, C.-M. Articles by Chiou, S.-Y. Search for Related Content PubMed PubMed Citation Articles by Chen, C.-M. Articles by Chiou, S.-Y.

Breast Lesions on Sonograms: Computer-aided Diagnosis with Nearly Setting-Independent Features and Artificial Neural Networks¹

¹ From the Institute of Biomedical Engineering, National Taiwan University, 1, Section 1, Jen-Ai Rd, Taipei 100, Taiwan (C.M.C., K.C.H., G.S.H.); Department of Radiology, Division of Ultrasound, Taipei Veterans General Hospital and National Yang Ming University, Taiwan (Y.H.C., C.M.T., H.J.C.); and Department of Radiology, Division of Ultrasound, Taipei Veterans General Hospital (S.Y.C.). Received November 19, 2001; revision requested January 28, 2002; revision received May 17; accepted June 27. Supported by National Science Council grant NSC90-2213-E-002-103, Taiwan. Address correspondence to C.M.C. (e-mail: ming@lotus.mc.ntu.edu.tw).

View larger version (124K): [in a new window]   Figure 1. Sonogram shows the inner gray contour as the lesion border and the outer white polygon as the convex hull of the lesion. Protuberances and depressions in the malignant breast lesion are indicated.

View larger version (25K): [in a new window]   Figure 2. Schematic shows lobes in a lesion, where {1, 2, 3, 4} are four representative concave points in and {A1, A2, A3, A4} are the sizes of the lobes.

View larger version (168K): [in a new window]   Figure 3. Sonogram shows the equivalent ellipse (thin line) of a malignant breast lesion, the boundary of which is marked by the thick line.

View larger version (168K): [in a new window]   Figure 4. Sonogram shows the skeleton of a malignant breast lesion. The boundary of the lesion is marked by the thick line, and the skeleton is indicated by the thin line segments.

View larger version (35K): [in a new window]   Figure 5. Two-layer feed-forward neural network used for classification. zi = the ith feature, vij = the weight of the synapse that connects the jth input to the ith neuron in the hidden layer, yi = the ith neuron in the hidden layer, wi = the weight of the synapse that connects the ith neuron in the hidden layer to the output, = the sigmoidal activation function, and o = the output.

View larger version (24K): [in a new window]   Figure 7. In bar graph, each pair of bars depicts the mean Az value and mean best classification accuracy achieved with a proposed morphologic feature when applied to four collections of lesion boundaries. Each collection comprised one of the four sets of lesion boundaries in the first set of US images and the lesion boundaries in the second set of US images. Error bars indicate 1 SD. At the 5% significance level, the first four features (ie, NSPD, LI, ENS, and ENC) are better than the other three. Moreover, the performances of NSPD, LI, and ENC are the same with the first set of US images and with all 271 images.

View larger version (24K): [in a new window]   Figure 6. In bar graph, each pair of bars illustrates the mean Az value and mean best classification accuracy achieved with a proposed morphologic feature with the first set of US images. Error bars indicate 1 SD. The first four features (ie, NSPD, LI, ENS, and ENC) are better than the other three at the 5% significance level.

View larger version (18K): [in a new window]   Figure 8. In bar graph, each pair of vertical bars indicates the mean Az value and mean best classification accuracy achieved with a Giger feature with the first set of US images. Error bars indicate 1 SD. All these features (NRG = normalized radial gradient) are worse than NSPD, ENS, and ENC at the 5% significance level.