Optimization in ultrasonic imaging systems: Use of metaheuristics to design sparse arrays and a heuristic to create an acquisiton system

• Ano de defesa: 2022
• Autor(a) principal: Souza, Júlio Cesar Eduardo de
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Estadual Paulista (Unesp)
• Palavras-chave em Português: Algoritmos de Otimização Estocástico; Arrays 2D Esparsos; Array Lineares Esparsos; Metaheurísticas; Linear Sparse Arrays; Metaheuristics; Stochastic Optimization Algorithms; 2D Sparse Arrays
• Link de acesso: http://hdl.handle.net/11449/237288

Ultrasonic images have an important contribution to medical diagnosis and nondestructive testing. One strategy to generate an image is to use an array, which is a transducer composed of a set of piezoelectric elements, that emits several wavefronts and samples the reflected waves. An important physical characteristic of arrays is their dimension. The wider the array extension, the better the lateral resolution of the generated image will be. Additionally, a construction recommendation for arrays is that the centre of their elements must be spaced by a distance (pitch) less or equal to 0.5λ, where λ is the generated wavelength by the transducer. Thus, the images generated by these arrays do not present artefacts caused by the grating lobes. The recommendation for using arrays is that it has to be the wider array possible, respecting the pitch recommendation. However, the data volume, resource, and manufacturing cost proportionally increase as the number of elements in the array rises, which might be impractical to use this array depending on the application. This thesis investigates techniques to reduce the use of resources in ultrasonic systems aiming to achieve images with high lateral resolution and mitigate any disadvantages. In the first part of this thesis, the linear sparse arrays, which are arrays that pitch higher than 0.5λ are studied. A new strategy to design these arrays is proposed, where a new mathematical codification for sparse arrays and fitness function based on the equation of energy and entropy of the PSFs (Point Spread Function) are presented. Subsequently, stochastic optimization algorithms are used to design sparse configurations. The proposed fitness function was compared with the most used fitness function in the literature based on the radiation pattern. The sparse arrays found using the proposed fitness function generated images with a balance between contrast and lateral resolution. Moreover, it was noticed that the fitness function proposed in the literature has inconsistencies when evaluating sparse array configurations which do not happen with the proposed fitness function. Next, a new data acquisition strategy for synthetic aperture for two-dimensional arrays that are not in a grid is proposed. This strategy is based on analyzing the projections of the elements of the coarray and keeping only the combinations of emitter and receiver elements that are most important for image generation. Consequently, the number of acquisitions and data volume of sparse two-dimensional apertures, whose elements are not positioned in a grid, is reduced, as well as the image generation time. The results indicate that it is possible to reduce the number of acquiring signals without compromising the quality of the ultrasonic image generated. In addition, two figures of merit based on the spatial distribution of the elements were used to evaluate sparse 2D arrays. A study of these parameters and how they influence the energy irradiated by arrays is done, and a fitness function is created. Then, a strategy to design a sparse 2D array is proposed using the simulated annealing algorithm. The radiation pattern analysis of the sparse arrays obtained from the search algorithm shown that the aperture generated images with high lateral resolution and low artefact intensities. The radiation pattern analysis of the sparse arrays obtained from the search algorithm showed that the aperture generated images with high lateral resolution and low artefact intensities. This thesis has three main contributions to ultrasonic systems that reduce manufacturing and computational costs.