אירועים והרצאות בפקולטה למדעי המחשב ע"ש הנרי ומרילין טאוב

The work addresses the two main challenges that have so far prevented adaption of chirp transmission in medical ultrasound, despite its potential to achieve a significantly enhanced SNR. The first is associated with the high sidelobe level resulting from signal compression using a matched filter. The second refers to the level of the sidelobes that is further increased since the pulse shape is distorted in the biological tissue in an unknown manner. This distortion is a result of the frequency dependent absorption in the living tissue. A high sidelobe level is also associated with the fact that due to the variability in the tissue absorption, the pulse shape itself is not known, so that even matched filtering results in a high level of errors. In this work we hypothesize that the signal can be modeled as a sum of a relative few 'strong reflectors' that are the main cause of sidelobes and a large number of relatively weak reflectors. Using this sparsity as a prior, we can solve for the position and (complex) reflectivity of the strong reflectors as well as for the precise form of the pulse shape. This allows the retrieval of the reflectivity function, applying the appropriate matched filter to the signal resulting from the weak reflectors. We propose to use a method based on the Least Angle Regression (LARS) algorithm for the solution of the non-linear optimization problem involved in the estimation of the positions and the (generally complex) amplitudes of the strong reflectors, as well as the precise pulse shape anywhere in the tissue.

Our basic assumption is that the reflectivity function is a sparse vector. Accordingly, LARS chooses a linear model to represent the data based on the correlation between the received and the transmitted signals. We have extended the LARS algorithm to allow its application to the case of complex vectors, which are often the case in practical ultrasound systems. The two major issues of the work are: (1) setting of a proper threshold to LARS and (2) addressing the problem of signal distortion caused by the frequency dependent attenuation of biological tissues. Since the threshold determines the number of entries in the decoded reflectivity function, while we deal with noisy environment, the tuning of the LARS' threshold has a major impact on the noise immunity, and on the rate of false detection.

Good estimation of the attenuation factor is achieved based on the correlation between the received and the transmitted signals. In addition, we have analyzed the attenuation influence on the decoded function and proposed a method to identify errors in the estimation. The proposed decoding method has been tested with real and complex reflectivity functions to evaluate the effects of attenuation and noise on the results. Our conclusion is that the proposed approach is superior to presently available methods in terms of detection in a noisy environment, and allows using chirps signals, with all their advantages in such systems.