Correlative feature analysis on FFDM
Authors:
- Yading YuanYYYading Yuan
- Maryellen L. GigerMGMaryellen L. Giger
- Hui LiHLHui Li
- Charlene SennettCSCharlene Sennett
- 13 November 2008
- journal article
- Published by Wiley in Medical Physics
- Vol. 35 (12) , 5490-5500
Abstract
Identifying the corresponding images of a lesion in different views is an essential step in improving the diagnostic ability of both radiologists and computer‐aided diagnosis (CAD) systems. Because of the nonrigidity of the breasts and the 2D projective property of mammograms, this task is not trivial. In this pilot study, we present a computerized framework that differentiates between corresponding images of the same lesion in different views and noncorresponding images, i.e., images of different lesions. A dual‐stage segmentation method, which employs an initial radial gradient index (RGI) based segmentation and an active contour model, is applied to extract mass lesions from the surrounding parenchyma. Then various lesion features are automatically extracted from each of the two views of each lesion to quantify the characteristics of density, size, texture and the neighborhood of the lesion, as well as its distance to the nipple. A two‐step scheme is employed to estimate the probability that the two lesion images from different mammographic views are of the same physical lesion. In the first step, a correspondence metric for each pairwise feature is estimated by a Bayesian artificial neural network (BANN). Then, these pairwise correspondence metrics are combined using another BANN to yield an overall probability of correspondence. Receiver operating characteristic (ROC) analysis was used to evaluate the performance of the individual features and the selected feature subset in the task of distinguishing corresponding pairs from noncorresponding pairs. Using a FFDM database with 123 corresponding image pairs and 82 noncorresponding pairs, thedistancefeature yielded an area under the ROC curve (AUC) ofwith leave‐one‐out (by physical lesion) evaluation, and the feature metric subset, which includeddistance, gradient texture, and ROI‐basedcorrelation, yielded an AUC of. The improvement by using multiple feature metrics was statistically significant compared to single feature performance.Keywords
Funding Information
- U.S. Public Health Service (CA89452)
- U.S. Public Health Service (P50‐CA125183)
This publication has 30 references indexed in Scilit:
- Cancer Statistics, 2008CA: A Cancer Journal for Clinicians, 2008
- Cancer Statistics, 2007CA: A Cancer Journal for Clinicians, 2007
- Mutual-information-based registration of medical images: a surveyIEEE Transactions on Medical Imaging, 2003
- A new Image Registration technique with free boundary constraints: application to mammographyComputer Vision and Image Understanding, 2003
- Active contours without edgesIEEE Transactions on Image Processing, 2001
- Computerized analysis of multiple-mammographic views: potential usefulness of special view mammograms in computer-aided diagnosisIEEE Transactions on Medical Imaging, 2001
- Ideal observer approximation using Bayesian classification neural networksIEEE Transactions on Medical Imaging, 2001
- Establishing the correspondence between control points in pairs of mammographic imagesIEEE Transactions on Image Processing, 1997
- Multimodality image registration by maximization of mutual informationIEEE Transactions on Medical Imaging, 1997
- Analysis of spiculation in the computerized classification of mammographic massesMedical Physics, 1995