Estimation of Speed of Sound Using Medical Ultrasound Image Deconvolution

In diagnostic ultrasound imaging the speed of sound is assumed to be 1540 m/s in soft tissues. When the actual speed is different, the mismatch can lead to distortions in the acquired images, and so reduce their clinical value. Therefore, the estimation of the true speed has been pursued not only because it enables image correction but also as a way of tissue characterisation. In this paper, we present a novel way to measure the average speed of sound concurrently with performing image enhancement by deconvolution. This simultaneous capability, based on a single acquisition of ultrasound data, has not been reported in previous publications. Our algorithm works by conducting non-blind deconvolution of an ultrasound image with point-spread functions based on different speeds of sound. Using a search strategy, we select the speed that produces the best-possible restoration. Deconvolution allows the use of ultrasound imaging machines almost without modification unlike most other estimation methods. A conventional handling of a transducer array is all that is required in the data acquisition part of our proposed method: the data can be collected freehand. We have tested our algorithm with simulations, in vitro phantoms with known and unknown speeds, and an in vivo scan. The estimation error was found to be +0.01 ± 0.60 % (mean ± standard deviation) for in vitro in-house phantoms whose speeds were also measured independently. In addition to the speed estimation, our method has also proved to be capable of simultaneously producing a better restoration of ultrasound images than deconvolution by an assumed speed of 1540 m/s, when this assumption is incorrect.