Reduced peak-hopping artifacts in ultrasonic strain estimation using the Viterbi algorithm

Abstract

Internal strain resulting from tissue deformation can be estimated by correlation processing of speckle patterns within complex (i.e., radio frequency) ultrasound images acquired during deformation. At large deformations, the magnitude of the correlation coefficient peak can be significantly lower than unity, so that random speckle correlations will exceed the true peak. This effect is called "peak hopping" and causes significant errors in displacement and deformation estimates. Here we investigate the Viterbi algorithm, a dynamic programming procedure, to overcome peak-hopping artifacts by finding the most likely sequence of hidden states in a sequence of observed events. It is well suited to motion estimation in elasticity-imaging studies because adjacent tissue elements remain adjacent following deformation. Particularly, tissue elements along an ultrasonic beam in one image lie along a 3-D continuous curve in the next image instant. The observed event in this case is the correlation coefficient of a pixel at a certain displacement. Radio-frequency data were generated before and after deformation with an average strain of 6%. Simulations were performed for a homogenous medium and for a medium with a stiffer inclusion. Results show that Viterbi processing of speckle-tracking outputs can significantly reduce peak-hopping artifacts.