De-Cluttering of Medical Ultrasound Images using Optimally-Compressed RF Data

Background: In medical ultrasound imaging, the use of transducer arrays improves the quality of the images, while adaptive signal processing in the aperture domain reduces the image distortion. Such a distortion is caused by the tissue inhomogeneity, which often results in clutter noise that impairs the diagnostic value of the image. For real-time imaging, special hardware systems are required in order to accommodate the processing of array signals. Software-based architectures provide benefits compared to hardware-based systems, however, their main bottle-neck lies in the data transfer load from the analog front-end to the digital back-end. Adaptive compression of RF data from the analog front-end may alleviate this limitation while improving the B-mode images quality.

Method: In order to compress the RF data, a transform representation is applied, where the appropriate quantization coefficients are chosen to meet a pre-defined data transfer rate while optimizing a measure in the rate-fidelity domain. This measure quantifies the fidelity of the B-mode image with respect to a de-cluttered ground-truth of the image. The de-cluttered image is chosen according to expert radiologist evaluation.

Results: The proposed algorithm was tested on echo-cardiac RF data, and yielded an approximate ten-fold (~10:1) compression, while producing de-cluttered images that resemble the ground-truth both visually and quantitatively.

Conclusions: The results demonstrate that the proposed algorithm enables to compress RF data whilst efficiently de-cluttering the produced B-mode images, therein improving their diagnostic quality. Our method thus has the potential to be implemented in real-time, software-based architectures for improved imaging systems performance.