| Ultrasound imaging has become one of the most widely used diagnostic tools in medicine. While it has advantages, compared with other modalities, in terms of safety, low-cost, accessibility, portability and capability of real-time imaging, it has limitations. One of the major disadvantages of ultrasound imaging is the relatively low image quality, especially the low signal-to-noise ratio (SNR) and the low spatial resolution. Part of this dissertation is dedicated to the development of digital ultrasound signal and image processing methods to improve ultrasound image quality.; Conventional B-mode ultrasound systems display the demodulated signals, i.e., the envelopes, in the images. In this dissertation, I introduce the envelope matched quadrature filtering (EMQF) technique, which is a novel demodulation technique generating optimal performance in envelope detection.; In ultrasonography, the echo signals are the results of the convolution of the pulses and the medium responses, and the finite pulse length is a major source of the degradation of the image resolution. Based on the more appropriate complex-valued medium response assumption rather than the real-valued assumption used by many researchers, a nonparametric iterative deconvolution method, the Least Squares method with Point Count regularization (LSPC), is proposed. This method was tested using simulated and experimental data, and has produced excellent results showing significant improvements in resolution.; During the past two decades, ultrasound tissue characterization (UTC) has emerged as an active research field and shown potentials of applications in a variety of clinical areas. Particularly interesting to me is a group of methods characterizing the scatterer spatial distribution. For resolvable regular structures, a deconvolution based method is proposed to estimate parameters characterizing such structures, including mean scatterer spacing, and has demonstrated superior performance when compared to conventional methods in situations of small data segments and highly randomized scatterer distribution. For non-resolvable structures, based on the idea of analyzing the distribution of large-valued signal points, several features, referred to as the DLP features, are extracted from the signals and used to characterize the scatterer number density (SND) of the tissues. The idea is further extended to characterize inhomogeneous tissues that are better represented by composite scatterer models and found successful as well.; A variety of physiological, histo-chemical, and morphological changes take place in skeletal muscles while aging. To apply the UTC techniques to study aging effects in skeletal muscles, in vitro experiments were made using extensor digitorum longus (primarily type II fibers), soleus (primarily type I), and quadriceps (mixed) muscles dissected from rats of three age groups (young, middle-aged, and old). Ultrasound RF signals were collected, and the attenuation coefficients and the DLP features were computed. Statistical analysis finds some significant differences between the features of different muscles in the same age group and between the features of the same muscles in different age groups. These findings are presented in chapter 5 and are consistent with the age-related changes previously observed and the behavior of the UTC features, showing the potential to develop UTC techniques as economical, easily accessible, and noninvasive tools for muscle condition evaluation. |
