Research On Beam Steering Techniques For Phased Medical Ultrasound Imaging

For ultrasound propagating through the organism, the character divergence ofbiological tissues results in the parameter variation of returned sound waves, such asamplitude, phase, and time. The returned waves are recorded, transformed, or be reverseprocessed to retrieve the tissue information. Thus, the images of the biological tissuescan be obtained. Because of its noninvasive, medical ultrasound imaging technology hasalready been widely accepted as one of the most important diagnostic methods inclinical.Phased array ultrasound imaging technology uses electronic focusing and scanningtechnologies to form beams with improved spatial and time characteristics in the wholemeasure area, which results reconstructed images with enhanced spatial resolution,extended detection dynamic range, and reduced geometry distortion. Thus, the imagingquality can be greatly improved compared with traditional ultrasound imaging methods.In this dissertation, the beamforming algorithms and its implementation details arethoroughly studied and analyzed. We also proposed three beamforming approaches，with different computation complexity, to improve the spatial resolution, time resolutionand uniformity of the images. Finally, a highly integrated, programmable array medicalultrasound imaging architecture consisting48independent channels, which can meet therequirements of medium-level ultrasound machines while providing a flexible platformfor supporting the development of new algorithms and emerging clinical applications ispresented. In summary, the main contributions of this dissertation are as follows:(1) The theory of ultrasound imaging and the model for calculating the field profilefrom arbitrarily shaped transducers under continuous wave and pulsed excitation arediscussed. The factors affecting the pressure fields and the improving methods are alsopresented. Finally, the imaging model based on phased pulse-echo fields is given. Thesimulation results show that the filed profile changes that were induced by scatterslocated in the front of transducer can be reconstructed by the returned waves, andenhanced image quality can be obtained by improving the directivity of the pulse-echofields. (2) The relationship between imaging dynamic range, spatial resolution and phasecontrol accuracy, focusing methods, beam directivity are discussed. A real-timebeamforming approach which can increase the fame-rate without scarifying imagingquality is proposed. This method uses1/4th transmit beams combined with four parallelreceive beams to increase the frame-rate by a factor of four. Mainlobe width controllingand sidelobes suppressing are also used to improve spatial resolution. Simulation resultsdemonstrate that improved imaging quality can be obtained comparing with traditionaldelay-and-sum bamforming methods.(3) A low computation complexity subspace-based adaptive bemaformingalgorithm is presented. In this method, Toeplitz matrix preprocessing and eigenvaluesreconstruction are employed to get a good estimation of array covariance matrix, whichis then employed in minimum variance (MV) weights calculation. Simulations on pointtargets and cyst demonstrate that the proposed bemaforming method can effectivelyimprove the performance of the beamformer in coherent interference environment. Theaperture loss results from spatial smoothing used by traditional beamformer can beavoided. Thus, the lateral resolution and contrast of beamforming images can besimultaneously improved.(4) The LS-SVM algorithm is extended for solving robust beamforming problems.In this approach, a squared-loss function is used to replace the conventional linearlyconstrained minimum variance cost function, which could significantly increaserobustness against mismatch problems and provide additional control over the sidelobelevel. Then, Gaussian kernels function is applied to the array observations to improvethe generalization capacity. Finally, a recursive regression procedure is presented to findthe solutions on real-time. Model reduction to reduce the final size of the beamformerwas also presented in the proposed approach. The test results show that the proposedbeamforming method significantly outperforms many other recently proposed linearrobust beamforming techniques in terms of signal distortion in the desired signal andnoise reduction in scenarios with DOA mismatch, limited observation samples, andnumerous interferences.(5) A programmable, highly integrated, phased array medical imaging architecture，which can support the development of new algorithms and emerging clinicalapplications is presented. This system can provide48independent transmit pulses. The pulse width and the number of transmit pulse can also be adjusted. The sample accuracyof the system is12bit and the sample rate can reach to50MHz. A two level variabletime delay technology is used to increase phase control accuracy of the transmit pulse to1.25ns at low system clock frequency. A fractional time delay polyphase filter based onleast square error is also provided to improve receving time resolution and realizedynamic focus. In addition, the plan of transmit pulse, the technology of scancontrolling, the method of mult-channel data synchronous sampling, and the design ofbeamformer are also discaussed with detail.As one part of phased array imaging technology, the research results in thisdissertation may lay a foundation for the development of beamforming algorithms andproviding some suggestions for the design of clinical applications from theoretical andpractical angle.