Diagnostic results for breast disease by real-time elasticity imaging system

Abstract

We have previously reported our real-time elasticity imaging system and a preliminary application to breast tissue diagnosis. We now propose several post processing algorithms to stabilize the elasticity imaging, introduce a method for scoring its clinical usefulness, and report the current status of the scoring method on breast tissue diagnosis. Our newly implemented post processing algorithms are: (1) frame-to-frame smoothing; (2) adaptive contrast optimization; (3) noisy-frame rejection; (4) noisy-region reduction. Using these algorithms, more than 20% of intensity fluctuations (SD) in strain images can be reduced. Our newly introduced scoring method is based on the imaging pattern of the low-strain region inside the hypoechoic region in the B-mode image. We classify 5 grades of elasticity score ranging from 1 (no strain-zero brightness region; benign) to 5 (broader strain-zero brightness region; malignant). As the result of applying 137 patients with breast diseases, this method provides a sensitivity of 87%, a specificity of 91%, and an accuracy of 89%.