Browsing by Subject "Contrast enhancement"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Post-beamforming filtering for enhanced contrast resolution in medical ultrasound.
    (2010-10) Wan, Yayun
    This thesis research addresses the use of post-beamforming pseudoinverse filtering algorithms for restoring contrast resolution in pulse-echo medical ultrasound imaging. Limited contrast resolution is probably the single most important limitation of ultrasound compared with leading medical imaging modalities. While speckle is a major contributor to the loss of contrast resolution, other significant contributors are low SNR and reverberations. We have investigated the use of coded excitation, together with a 1D post-beamforming pseudoinverse filter, in improving the SNR and reducing reverberation leading to enhanced contrast resolution, especially for deeper target. Experimental demonstration of the algorithm was carried out on a dual-mode ultrasound array (DMUA) prototype intended for use in image-guided noninvasive surgery. We have successfully used coded excitation with receive pseudoinverse filtering to improve the resolution about 30% while maintaining the signal to noise ratio, reducing reverberation artifact. In addition, this result in conjunction with other techniques such as spatial directivity and gain compensation significantly improve the imaging field of view of the DMUA system. With the advent of digital beamforming in array imaging, coarse aperture sampling is identified as another significant contributor to the loss of contrast resolution, especially in high frequency ultrasound (HFUS) imaging applications. This loss results from well-known beamforming artifacts (e.g. grating lobes), which produce "filling effects" in low-contrast targets (e.g. cysts and blood vessels). To address this issue, a 2D post-beamforming filtering approach was formulated from a discretized model of the transmit-receive 2D wavefront resulting from a given beamforming operation. This 2D filter operates on a collection of beamformed A-lines (e.g. from a linear array) with coefficients obtained from the regularized inversion of the 2D Fourier transform of the 2D point spread function of the array. This is highly significant due to the fact that direct inversion of the imaging matrix for a typical HFUS imaging scenario requires on the order of 100T flops. This was enabled by deriving the imaging operator on a 2D Cartesian grid which, under realistic simplifying assumptions, was shown to be represented by a matrix with a Toeplitz-block block Toeplitz (TBBT) structure. The dimensions of the TBBT are extremely large, which renders the direct inversion impractical, both in terms of memory requirements and number of operations. However, the large TBBT matrix has an asymptotically equivalent Circulant-block block Circulant (CBBC) matrix with equivalent eigenvalues. The CBBC is easily inverted using a finite-size 2D discrete Fourier Transform. Not only is this approach computationally efficient, it also results in a robust, physically meaningful regularized inversion. In particular, the approach transforms a usually ill-posed inverse problem to a well-posed filtering problem in k-space (through a 2D FFT). An important result from this study is that the spatial and contrast resolutions vary monotonically with the regularization parameter. This result is of practical significance as it allows for the selection of the optimal value of the regularization parameter in much the same way as time gain compensation, e.g. slider or dial. Using FIELD II, we present simulation data to demonstrate the tradeoff between contrast and spatial resolution. The results demonstrate the well-behaved nature of the point spread function (PSF) with the variation in a single regularization parameter. This characteristic of the pseudoinverse filter enables a parameter-controlled and more importantly, user-controlled imaging performance. These results are supported by image reconstructions from a simulated cyst phantom obtained using a finely sampled array and a coarsely sampled array. These results are also verified by image reconstructions obtained from Sonix RP system imaging a quality assurance phantom with contrast targets, optical nerve head in porcine eye in vitro and human carotid artery in vivo.