Ultrasound (US) elastographymeasures the deformation of tissue in response to applied force to derive anddisplay its stiffness. It has been applied to various organs and tissues, e.g.,as an adjunct tool to conventional US and X-ray mammography in the detectionand characterization of breast masses. Real-time US elastography processing isdemanding due to several computationally-expensive modules, such ascrosscorrelation. Multi-core digital signal processors (DSPs) provide increasedcomputational performance with less power consumption and lower cost than otherprogrammable approaches, e.g., with general-purpose CPUs and GPUs. However,optimal algorithm implementations on multi-core processors require carefulalgorithm/architecture analysis and prudent mapping to achieve goodperformance. In this paper, we present our approach and techniques used inmapping our US elastography algorithm on a multi-core DSP, i.e., TexasInstruments TMS320C6472. Our results show that a single C6472 multi-core DSPcan support real-time US elastography processing.
|