Real-time ultrasound elastography

Ultrasound elastography can provide tissue stiffness information that is complementary to the anatomy and blood flow information offered by conventional ultrasound machines, but it is computationally challenging due to many time-consuming modules and a large amount of data. To facilitate real-time implementations of ultrasound elastography, we have developed new methods that can significantly reduce the computational burden of common processing components in ultrasound elastography, such as the crosscorrelation analysis and spatial filtering applied to displacement and strain estimates. Using the new correlation-based search algorithm, the computational requirement of correlation-based search does not increase with the correlation window size. For typical parameters used in ultrasound elastography, the computation in correlation-based search can be reduced by a factor of more than 30. Median filtering is often performed to suppress the spike-like noise that results from correlation-based search. For fast median filtering, we have developed a method that efficiently finds a new median value utilizing the sort result of the previous pixel. With careful mapping of the new algorithms on digital signal processors, our work has led to development of a clinical ultrasound machine supporting real-time elastography. Our methods can help real-time implementations of various applications including ultrasound elastography, which could lead to increased use of ultrasound elastography in the clinic.