The real-time nature of medical ultrasound, combined with the ease of use, the low cost and the portability of its equipment renders ultrasound imaging an indispensable tool for both diagnostic and guided applications. Ultrasound has the ability to penetrate deep into soft tissues maintaining a good spatial resolution. Despite the great advances in ultrasound imaging technology, there is still space for improvement in terms of image resolution enhancement. This research focuses on the algorithmic development of signal processing and system identification techniques that tackle the unwanted effects of noise originating either from the body itself or the ultrasound system, providing higher quality images with enhanced diagnostic value.

Medical ultrasound is a noninvasive imaging technique that uses high-frequency sound waves to create real-time images of the human body. The procedure involves an ultrasonic transducer that transmits short acoustic pulses and records the echoes that are reflected back from internal structures of targeted body regions. These echoes are then used to generate fine images of the underlying anatomy.

The interaction between the acoustic pulse generated by the transducer and the scanned human tissue introduces a particular type of blurring to the output image (as shown in the figure below), reducing both contrast and resolution. The removal of the pulse from the ultrasound image is not an easy task since tissue heterogeneity induces pulse shape distortions. Furthermore, image quality is degraded by the presence of noise that arises either from the electronics of the detection system or the large variations in the acoustic properties of the body. One such example is reverberation noise, i.e. echoes that are reflected back and forth several times before they reach the transducer.

The main goal of this research is the development of compensation strategies that reduce the impact of acoustic and anatomic artifacts, allowing proper image interpretation. Specifically, we focus on eliminating the blurring effect of the transmitted pulse on the obtained images. To this end, we have developed data-driven algorithms that can accurately model the transmitted pulse and track changes in its shape down through the tissue. By removing the distorted pulse from the ultrasound image, we achieve improved contrast and image resolution. Compared to other techniques, the aforementioned data-driven methodology is fully automated and requires no prior knowledge of the underlying system characteristics. Regarding the suppression of noise artifacts, originating either from the body or the ultrasound system, we combine adaptive filtering methods and time-varying system identification techniques taking into account the exact position and the shape of the artifacts, as well as prior information about the anatomy.