Study On Adaptive Beamforming Method For Medical Ultrasound Imaging

Ultrasound imaging has been widely used in clinical diagnosis for all kinds of human tissues. The main limit for it is the low spatial resolution and image contrast, compared with other medical imaging modalities, which brings difficulties to diagnosis and image processing. Beamforming is the most critical part that determines ultrasound image quality, on which research is of great importance to the further promotion of clinical ultrasound imaging.The conventional delay and sum (DAS) beamformer is simple to implement, but its broad mainlobe and high sidelobes result in poor image quality. Since the21st century, adaptive beamformers have been developed for medical ultrasound imaging. While improving image quality to some extent, they meet their own limitations. Based on medical ultrasound beamforming theories, three adaptive algorithms have been proposed to enhance current methods.The most commonly used minimum variance (MV) beamformer significantly improves the lateral resolution. However, it suffers from signal cancellation in the presence of highly correlated interference, which degrades the image contrast. To overcome this drawback, an iterative minimum variance (IMV) beamformer and a spatio-temporal amplitude and phase estimation (ST-APES) beamformer are proposed. Based on the conventional MV beamformer,, the IMV beamformer iteratively estimates the interference-plus-noise covariance matrix to calculate a set of new weights. The ST-APES beamformer extends the APES algorithm to the spatial domain, and estimates the true signal on the least square criterion in both spatial and temporal dimensions, combined with the a priori information of transmitted pulse waveform. Moreover, the high-contrast eigenspace-based MV (EIBMV) beamformer tends to distort the speckle background around strong scatterers. Then, a modified EIBMV beamformer and a region detection-based eigenspace MV (RD-ESMV) method are proposed. The RD-ESMV method detects ultrasonic hypoechoic regions according to the relationship between the eigenvectors and the steering vector, along with the eigenvalues, and then implements different bqamformers for different types of regions.Simulations and experiments have been conducted for point scatterers, anechoic cyst, massive cyst and situations with sound speed errors. Results show that the IMV and ST-APES beamformer both reduce signal cancellation, and achieve more accurate amplitude estimation and higher image contrast than the conventional MV. The ST-APES beamformer is also more robust against sound speed errors. The RD-ESMV method effectively avoids speckle distortions, and simultaneously maintains the high contrast of the EIBMV beamformer. Hopefully, it will become a more practical beamforming approach in applications where speckle pattern and texture information are important.