Sistema ETD

Archivio digitale delle tesi discusse presso l'Università di Pisa


Tesi etd-02272014-135351

Tipo di tesi
Tesi di dottorato di ricerca
Development and analysis of Magnetic Resonance Imaging acquisition and reconstruction methods for functional and structural investigation of cardiac and lung tissues.
Settore scientifico disciplinare
Corso di studi
commissario Bianchi, Anna M.
commissario Wiesinger, Florian
tutor Prof. Landini, Luigi
commissario Prof.ssa Laschi, Cecilia
Parole chiave
  • non-Cartesian acquisition
  • navigators.
  • lung
  • inner volume
  • geometric information
  • cardiac
  • reconstruction
  • T2 mapping
Data inizio appello
Riassunto analitico
The imaging of the lung and of the heart are often challenging in magnetic resonance due to the motion of the organs. In order to avoid motion artifacts it is possible to make the acquisition fast enough to fit in the breath-hold, or use some motion management methods in free breathing.
A fast image acquisition can be obtained with non-Cartesian acquisition schemes, which require specialized reconstruction methods. In this work the least-squares non-uniform fast Fourier transform (LS-NUFFT) was compared to the standard gridding (GR) taking the direct summation method (DS) as reference. LS-NUFFT obtained lower root mean square error (RMSE), but heavier geometric information loss. The performance improvement of the LS-NUFFT was studied using three regularization methods. The truncated SVD reduced the RMSE of the simple regularization-free LS-NUFFT.
Alternatively, the scan time can be shortened with some FOV reduction techniques. For cardiac imaging, the inner volume (IV) reduced-FOV selection was explored for the myocardial T2 mapping. The FOV reduction successfully avoided aliasing and provided a scan time reduction from about 23s to 15s. However, undesired stimulated echoes caused an overestimation in the T2 of about 20%. The effects of the inner volume excitation on the T2 mapping were described and clarified.
Finally, motion management was explored for lung imaging in free-breathing, using a non-Cartesian acquisition trajectory. The rotating ultra-fast sequence (RUFIS) was demonstrated to be very suitable for the short T2* lung tissue. The respiratory motion was addressed with three methods: prospective triggering (PT), prospective gating (PG) and retrospective gating (RG). All methods were able to reconstruct a 3D high-resolution dataset. PG and RG could achieve 1.2 mm isotropic resolution in clinically reasonable scan time (~6min). The RG sequence could reconstruct multiple phases of the respiration cycle at cost of higher scan time.