Shin, Bonghun2017-10-062017-10-062017-10-062017-10-04http://hdl.handle.net/10012/12534Portable wireless ultrasound is emerging as a new ultrasound device due to the advantages such as small size, lightweight and affordable price. Its high portability allows practitioners to make diagnostic and therapeutic decisions in real-time without having to take the patients out of their environment. Recent portable ultrasound devices are equipped with sophisticated processors and image processing algorithms providing high image quality. Some of them are able to deliver multiple ultrasound modes including color Doppler, echocardiography, and endovaginal examination. Nevertheless, they are still lack of elastography functions due to the limitations in computational performance and data transfer speed via wireless communication. In order to implement the elastography function in the wireless portable ultrasound devices, this thesis proposes a new strain estimation method to significantly reduce the computation time and a compressive sensing framework to minimize the data transfer size. Firstly, a robust phase-based strain estimator (RPSE) is developed to overcome the limited hardware performance of portable ultrasound. The RPSE is not only computationally efficient but also robust to variations of the speed of sound, sampling frequency and pulse repetition. The RPSE has been compared with other representative strain estimators including time-delay, displacement-gradient, and conventional phase-based strain estimators (TSE, DSE and PSE, respectively). It has been shown that the RPSE is superior in several elastographic image quality measures, including signal-to-noise (SNRe) and contrast-to-noise (CNRe), and the computational efficiency. The study indicates that the RPSE method can deliver the acceptable level of elastography and fast computational speed for the ultrasound echo data sets from the numerical and experimental phantoms. According to the results from the numerical phantom experiment, RPSE can achieve highest values of SNRe and CNRe (around 5.22 and 47.62 dB) among all strain estimators tested, and almost 100 times higher computational efficiency than TSE and DSE (around 0.06 vs. 5.76 seconds per frame for RPSE and TSE, respectively). Secondly, as a means to reduce the large amount of ultrasound measurement data that has to be transmitted via wireless communication, the compressive sensing (CS) framework has been applied to elastography. The performance of CS is highly dependent on the selection of model basis to represent the sparse expansion as well as the reconstruction algorithm to recover the original data from the compressed signal. Therefore, it is essential to compose the optimal combination of model basis and reconstruction algorithm for CS framework to achieve the best CS performance in terms of image quality and the maximum data reduction. In this thesis, three model bases, discrete Fourier transform (FT), discrete cosine transform (DCT), and wave atoms (WA), along with two reconstruction algorithms, L1 minimization (L1) and Block sparse Bayesian learning (BSBL) are tested. Using B-mode and elastogram images of simulated numerical phantoms, the quality of CS reconstruction is assessed in terms of three image quality measures, mean absolute error (MAE), SNRe, and CNRe, at varying data reduction (subsampling) rates. The results illustrate that BSBL based CS frameworks can generally deliver much higher image quality and subsampling rate compared with L1-based ones. In particular, the CS frameworks adopting DCT and BSBL offer the best CS performance. The results also suggests that the maximum subsampling rates without causing image degradation are 40% for L1-based framework and 60% for BSBL-based framework, respectively. The contributions of this thesis help realize elastography functionality in portable ultrasound, thereby significantly expanding its utility. For example, the diagnosis of malignant lesions, even when a patient cannot be moved to hospital immediately, is possible with the portable ultrasound. Furthermore, the SPSE method and the CS framework can be individually employed for the conventional ultrasound device as well as other telemedicine applications, to enhance computational efficiency and image quality.enElastographyStrain ImagingUltrasoundTelemedicineTime-Delay Strain EstimationPortable Ultrasound DevicesPhase-Based Strain EstimationCompressive SensingSparse Bayesian LearningL1 MinimizationOptimizationDoctoral Thesis