polito.it
Politecnico di Torino (logo)

Computationally Optimized 3D Stack-of-Spirals Image Reconstruction

Zoia Laraib

Computationally Optimized 3D Stack-of-Spirals Image Reconstruction.

Rel. Danilo Demarchi, Benedikt Poser. Politecnico di Torino, Corso di laurea magistrale in Ict For Smart Societies (Ict Per La Società Del Futuro), 2023

[img]
Preview
PDF (Tesi_di_laurea) - Tesi
Licenza: Creative Commons Attribution Non-commercial No Derivatives.

Download (1MB) | Preview
Abstract:

Spiral readout trajectories have become popular in several applications of 2D and 3D MR imaging applications because of their ability to acquire more data in less time than their traditional Cartesian counterparts. 3D sampling is especially desirable for high-resolution imaging, which results in the combined burden of the (i) complexity inherent to non-Cartesian reconstruction, (ii) large data matrices for high resolution, and (iii) the need to process the entire 3D dataset in one go (instead of slice-by-slice as in 2D). The consequences are high computational cost and memory requirements during the reconstruction process, especially in combination with parallel imaging and off-resonance correction. This research aims to improve the computational efficiency of the reconstruction methods while maintaining the accuracy of the resulting 3D images. Several challenges need to be addressed to improve the accuracy and efficiency of 3D imaging, such as incomplete k-space coverage, motion artifacts, noise, and field inhomogeneities. A four-step approach is proposed to optimize the 3D spiral reconstruction process for high??resolution single-shot under-sampled spirals and with correction for magnetic field inhomogeneities. Optimal k-space coverage and artifact minimization is critical for selecting an appropriate spiral trajectory. Iterative reconstruction methods have been widely implemented for accurate and efficient image reconstruction, but optimization of these methods, especially 3D, can be time-consuming and computationally expensive. This work uses the high-level programming language Julia, which recently emerged as a powerful tool for implementing iterative reconstruction methods due to its speed and built-in support for parallel computing. By combining Julia with iterative reconstruction methods, we find that the efficiency and accuracy of MR image reconstruction can be greatly enhanced: Julia's multi-threading and distributed computing support can significantly reduce the reconstruction time for large datasets. At the same time, its built-in libraries for linear algebra and optimization can simplify the implementation of complex algorithms. This is exploited in the computational optimization strategies and image quality in 3D imaging using spiral trajectories, focusing on reducing reconstruction time, minimizing the g??factor, and using pseudo-replica methods to analyze image variability. The optimizations proposed in this work are an important step towards achieving real-time high-quality 3D spiral reconstruction with a reduced computational burden. This will ultimately help applications in neuroimaging or clinical applications.

Relatori: Danilo Demarchi, Benedikt Poser
Anno accademico: 2022/23
Tipo di pubblicazione: Elettronica
Numero di pagine: 47
Soggetti:
Corso di laurea: Corso di laurea magistrale in Ict For Smart Societies (Ict Per La Società Del Futuro)
Classe di laurea: Nuovo ordinamento > Laurea magistrale > LM-27 - INGEGNERIA DELLE TELECOMUNICAZIONI
Aziende collaboratrici: NON SPECIFICATO
URI: http://webthesis.biblio.polito.it/id/eprint/26928
Modifica (riservato agli operatori) Modifica (riservato agli operatori)