Resumen de Métodos de reconstrucción en dominio temporal para tomografía por transmisión de ultrasonidos
Breast cancer (BC) is the leading cause of cancer-related death for women in Europe, and the second one after lung cancer in the US. Early detection is very important for the survival rate of BC, because the smaller the local extension of the neoplasia, the better the output of the surgical treatments employed and the lower the probability of needing more invasive treatments. Mammography is currently the standard procedure employed for breast screening programs around the world. Nevertheless, its efficiency has been questioned lately because: (i) it generates many abnormal findings not related to cancer, (ii) it requires irradiating the patient and (iii) it has low specificity with dense breasts. Consequently, complementary techniques to mammography are being proposed to improve the detection and characterization of BC. Among these techniques, is Ultrasound Computed Tomography (USCT). The transmission mode of USCT provides quantitative maps of sound speed (SS) and acoustic attenuation (AA) of tissue, which have been proposed for BC detection as they can improve the detectability of malignancies in the breast.
In this thesis, the main goal was to set up the transmission modality of USCT in the scanner MUBI, which was created as part of the collaboration of the Ultrasound Systems and Technology Group of the CSIC (USTG-CSIC), and the Nuclear Physics Group of the UCM (GFN-UCM), under the project TOPUS [S2013/MIT-3024, 2013-2017]. Methods for data processing and image reconstruction techniques were developed and implemented to this end. In first place, we implemented three approximate and fast geometrical acoustics (GAc) methods to obtain both SS and AA maps. These methods are based on, in one hand, the use of straight rays (analytical method based on the Radon transform) and on the other hand, on the use of bent rays (iterative methods that take into account the refraction experienced by the waves when crossing regions with different acoustic impedance). We implemented a classical bent rays method based on the Fast Marching Method (FMM) and compared its performance against a simple, fast and straightforward method for bent rays based on the Bézier polynomials, firstly proposed in this thesis. Several algorithms to obtain time of flight and amplitude of the signals, needed by GAc algorithms, were developed and tasted. Additionally, we investigated ways to improve the image quality by using full wave inversion (FWI) reconstruction. A framework in time domain to jointly perform the reconstructions of both SS and AA was proposed in this thesis. The developed methods were tested first with synthetic data and then they were applied to the reconstruction of real data obtained from the MUBI scanner using tissue mimicking phantoms.
The analytical method studied is able to give real-time information of the inspected regions without providing very high resolution. This information can be improved with our Bézier-based bent-rays method during the post-processing of the data for SS reconstructions. It provides faster and higher quality results in comparison with our implementation of the classical FMM. However, not considerable improvements were observed for the reconstruction of the AA with this method. GAc algorithms in general are not suitable to reconstruct high-quality AA maps. In case that a suspicious region is observed (very high SS), complimentary information can be obtained from the FWI reconstruction of the AA, as this method provides high-quality reconstructions of both SS and AA properties. Further experimental tests are still required to fully validate our reconstruction algorithms. Nevertheless, either with the ray based methods and the FWI, our algorithms have been tested with both synthetic and experimental data, showing very encouraging results in terms of quality and reconstruction times. We expect these results could be useful to expand and improve this promising technique.
