Algorithms for ultrasonic signal and image processing for diagnosis and mini invasive therapies

Due to their non invasive features, ultrasound (US) techniques are receiving an increasing attention in modern medical procedures. However, the results of diagnostics using conventional B-mode images are often difficult to interpret, since the US image quality is degraded by speckle noise due to random interference of the returns issued from numerous sub-resolution scatterers present in the tissue. For this reason, there is a strong need of automating the process of finding sick tissues in order to reduce the gap between skilled and novice physicians in using ultrasound as a diagnostics tool or to improve biopsy guidance and therapy treatment targeting.
The goal of this research line is the development of time-frequency and time- scale based algorithms to directly process RF biomedical signals in order to identify different tissue characteristics and their particular property such as impedance, attenuation, elasticity, or sound velocity. The investigation of different statistical approaches to describe echo signals and diffusion models, considering both non-uniform phase distribution and echo intensity, would allow us to parameterize regions of ultrasound B-scan images in terms of the number of scatterers in suitable local image areas (resolution cells) and evaluate coherent versus diffuse scattering components. Estimation of coherent and diffuse reflectors represents a powerful tool for speckle identification, to aid segmentation and tissue characterization. Speckle identification is particularly useful, since it is required for adaptive noise suppression algorithms and for use in decorrelation algorithms to estimate the elevation distance between neighboring B-scans. Finally, the design of electronic systems can be planned for real-time denoising and classification operations up to the working frequency and bandwidth constraints of modern echographic platforms.
This research has been developed within the context of PRIN and FIRB projects, in collaboration with other technical (Univ. Firenze) and medical (Univ. Siena) partners.