DEVELOPMENT AND IMPLEMENTATION OF INNOVATIVE ALGORITHMS FOR HIGH FRAME RATE ULTRASOUND

High-frame-rate (HFR) ultrasound imaging has revolutionized the way dynamic phenomena in tissues and blood flow are observed, enabling the reconstruction of thousands of images per second through unfocused transmissions such as plane or diverging waves. However, the real-time implementation of advanced HFR techniques, such as Color Flow Mapping (CFM), Power Doppler (PD), and Vector Doppler Imaging (VDI), requires high computational power and efficient data handling, which are difficult to achieve with conventional ultrasound systems. The aim of this Ph.D. work was the development of a heterogeneous ultrasound platform capable of combining the advantages of hardware-oriented systems (based on FPGAs and DSPs) and software-oriented systems (based on GPUs), in order to enable the real-time implementation of complex ultrasound imaging algorithms. To this end, a GPU module was integrated into the hardware-oriented ULA-OP 256 research system, developed at the MSDLab of the University of Florence, resulting in the heterogeneous system ULA-OP 256G. As a case study, the system was first validated by implementing the HFR CFM method on the GPU, which exhibited real-time performance superior to DSP-based implementations. Subsequently, the Vector Doppler Imaging method was implemented in real time on the ULA-OP 256G platform, enabling the reconstruction of bidimensional velocity fields by combining information from two different observation directions. In addition, an alias-free Doppler technique based on staggered pulse repetition frequency sequences (Staggered Pulse Repetition Frequency, S-PRF) was implemented, extending the Nyquist limit and correcting aliasing artifacts in real time. Finally, a real-time Singular Value Decomposition (SVD) filtering framework was developed to improve clutter suppression and enhance sensitivity to weak blood signals, demonstrating the feasibility of advanced spatiotemporal filtering directly on the embedded GPU. Experimental validations, performed on flow phantoms using linear probes, confirmed the ability of the developed methods to produce accurate and alias-free velocity maps. The proposed solutions demonstrate the potential of heterogeneous architectures for advanced ultrasound imaging and represent a significant step toward the clinical translation of real-time HFR Doppler techniques.