A modified block matching method for real-time freehand strain imaging.

Abstract

This manuscript reports a technical innovation that has been developed for real-time, freehand strain imaging. This work is based on a well-known block-matching algorithm with two significant modifications. First, since displacements are estimated row-by-row, displacement estimates from the previous row are used to predict the displacement estimates in the current row, thereby drastically reducing the search-region size and increasing computational efficiency. Second, a displacement error detection and correction method is developed to overcome the local tracking errors that may be more severe with freehand scanning and thereby improve the robustness of the algorithm. This algorithm has been implemented on a clinical ultrasound imaging system, and with real-time imaging feedback, long sequences of high quality strain images are observed using freehand compression. Displacement estimation errors with this method are experimentally measured and compared with results from simulation. We report only a specific implementation, with no comparison to other displacement estimators in the literature and no optimization of this specific technique. Images of tissue-mimicking phantoms with small spherical targets are used to test the ability to detect small lesions using the strain imaging technique. In vivo strain images of breast and thyroid are also shown.