IOP PUBLISHING PHYSICS IN MEDICINE AND BIOLOGY Phys. Med. Biol. 53 (2008) 4809–4823 doi:10.1088/0031-9155/53/17/022 Breast tumor classification using axial shear strain elastography: a feasibility study Arun Thitaikumar 1,3 , Louise M Mobbs 2 , Christina M Kraemer-Chant 2 , Brian S Garra 2 and Jonathan Ophir 1 1 Department of Diagnostic and Interventional Imaging, Ultrasonics Laboratory, The University of Texas Medical School, 6431 Fannin St, Houston, TX 77030, USA 2 Department of Radiology, University of Vermont college of Medicine, Burlington, VT, USA E-mail: Jonathan.Ophir@uth.tmc.edu Received 15 November 2007, in final form 17 June 2008 Published 13 August 2008 Online at stacks.iop.org/PMB/53/4809 Abstract Recently, the feasibility of visualizing the characteristics of bonding at an inclusion-background boundary using axial-shear strain elastography was demonstrated. In this paper, we report a feasibility study on the utility of the axial-shear strain elastograms in the classification of in vivo breast tumor as being benign or malignant. The study was performed using data sets obtained from 15 benign and 15 malignant cases that were biopsy proven. A total of three independent observers were trained, and their services were utilized for the study. A total of 9 cases were used as training set and the remaining cases were used as testing set. The feature from the axial-shear strain elastogram, namely, the area of the axial-shear region, was extracted by the observers. The observers also outlined the tumor area on the corresponding sonogram, which was used to normalize the area of the axial-shear strain region. There are several observations that can be drawn from the results. First, the result indicates that the observers consistently (∼82% of the cases) noticed the characteristic pattern of the axial-shear strain distribution data as predicted in the previous simulation studies, i.e. alternating regions of positive and negative axial-shear strain values around the tumor–background interface. Second, the analysis of the result suggests that in approximately 57% of the cases in which the observers did not visualize tumor in the sonogram, the elastograms helped them to locate the tumor. Finally, the analysis of the result suggests that for the discriminant feature value of 0.46, the number of unnecessary biopsies could be reduced by 56.3% without compromising on sensitivity and on negative predictive value (NPV). Based on the results in this study, feature values greater than 0.75 appear to be indicative of malignancy, while values less than 0.46 to be indicative of 3 Present address: University of Texas-M.D. Anderson Cancer Center, Houston, TX, USA. 0031-9155/08/174809+15$30.00 © 2008 Institute of Physics and Engineering in Medicine Printed in the UK 4809