Research On Adaptive Beamforming Methods In Medical Ultrasound Imaging

As one of the most often used medical imaging modalities, ultrasound imaging is a safe, portable, real-time and low-cost modality and has been widely used in clinical diagnosis. Beamforming lies in the center of an ultrasound system and plays a critical role in improving the image quality. The conventional delay-and-sum (DAS) beamforming, commonly used in commercial products, can be implemented simply and effectively for ultrasound image formation. However, it produces beams with a wide main lobe and high side lobe levels, resulting in images with low spatial resolution and contrast. The traditional apodization function uses predetermined weights (e.g., Hamming window) to reduce the side lobes of the DAS beamformer, yet broaden the main lobe, i.e., at the expense of resolution. In contrast to the non-adaptive blind DAS, adaptive beamforming technologies generate high-quality beams with narrower main lobe width and lower side lobes by computing aperture weights using the characteristics extracted from the received signals. Therefore, they can improve image resolution and contrast simultaneously. Here, we focus on coherence-based-factor adaptive weighting and minimum variance (MV) beamforming. Their imaging benefits and shortcomings are analyzed, and the corresponding methods of improvement are proposed as follows.First, the coherence factor (CF) and the phase/sign coherence factor (PCF/SCF) can suppress side and/or grating lobes and reduce clutter significantly with low computational complexity. However, they may produce negative effects, such as decreased average image brightness, increased speckle variance and even some loss of texture information. In this thesis, a phasor dispersion based coherence factor (PDCF) is developed using complex analytic signals of aperture data. By means of weighting the coherent beam sum, it is used to suppress side lobes and random noise and reduce image clutter. Experiments of simulated and real data sets indicate that compared with the DAS beamformer, PDCF demonstrates significant improvement in resolution and contrast. However, it may result in the defects in terms of speckle pattern. To solve the problem, a user-defined parameter a can be adjusted to compromise between suppressing off-axis interference and maintaining image texture information. The results show that the value of a slightly less than1, such as between0.9and1, can present a preferable trade-off.Second, adaptive MV-based beamformer has shown high-resolution property in ultrasound imaging, but its success in suppressing side lobes and enhancing the contrast has not yet been satisfactory. To this end, the combination between MV beamforming and phase coherence imaging is proposed. The results with simulated point and cyst phantoms demonstrate that this method can reduce sidelobes and clutter while retaining high resolution. Unfortunately, it also brings the problems such as reduced overall brightness and increased speckle variance.Third, the deficiencies occurred in the images formed by coherence-based methods, such as CF, PCF/SCF, are described in detail. They include reduced overall brightness, increased speckle variance, black-region artifacts around hyperechoic reflectors and underestimated magnitudes of point targets in large depth. In this thesis, the reasons for these artifacts are analyzed theoretically, and a new spatio-temporal smoothing procedure is proposed to rectify the coherence-based methods. The resulting spatio-temporally smoothed coherence factor (StS-CF) measures the signal coherence among the beamsums of the divided overlapping subarrays over the duration of a transmit pulse. Simulated and real experimental data sets demonstrate that the proposed method can improve the robustness of coherence-based factors, reduce speckle variance and remove black-region artifacts significantly, while preserving the capability of clutter suppression. Consequently, the detectability of cyst-like structures and the contrast can be enhanced.Finally, three approaches based on the StS-CF are proposed to further improve the imaging quality, including spatial filtering of the coherence factor, combining with spatial compounding technique and combining with the MV beamformer. Simulation results indicate the expected performance of the proposed methods. Spatial filtering of the coherence factor can reduce the speckle intensity variations resulted from the weighting fluctuation, remove black-region artifacts and improve the contrast. Combining with the spatial compounding can achieve the benefits of both, yielding preferable images with fewer artifacts, more homogeneous background and significantly improved contrast due to the suppression of both speckle noise and clutter. Combining with the MV beamformer can lead to simultaneous improvement of the resolution and contrast, while retaining background texture or speckle patterns. The overall image quality can be enhanced therefore.In summary, the above improvements of coherence-based factor and MV beamforming compensate for the shortcomings, increase the imaging performance and make them more suitable for clinic applications. In particular, the proposed StS-CF provides an instructive attempt to improve image quality and diagnostic capability of ultrasound systems.