This Thesis focuses on the development of advanced solutions for electromagnetic recovery and synthesis problems relevant for Magnetic Resonance Imaging (MRI). Two open issues in MRI are: 1) the exploitation of this imaging modality to theoretically quantify physical, chemical, or physiological property to be useful for different biomedical application, f.i. therapeutic treatments and diagnosis; 2) the necessity to obtain high resolution images, minimizing any kind of artifacts that may appear due to the high field level employed during an MRI scan. As far as the first issue is concerned, a possibility is the use of the radiofrequency (RF) field collected inside the MRI scanner to non-invasively retrieve the in-vivo electrical properties of biological tissues, that is the so-called MRI-based Electrical Properties Tomography. On the other hand, an essential parameter that can improve to enhance the image quality is the homogeneity of the RF field, referred as RF shimming. Results presented in this Thesis concern new physics-assisted approaches developed for both MRI-EPT and RF shimming. In this respect, we present novel learning-based methodologies for MRI-EPT, and, on the other hand, a new RF shimming procedure that addresses the underlying issue as a field shaping problem. Unlike most of the learning approaches that adopt neural networks as “black-boxes”, the proposed learning-based strategies, take into account the physics of the problem provided by the computation of the gradient. As far as the shimming procedure, it is further optimized with the development of an auxiliary model for the convenient selection of the parameter to be employed in the shaping optimization. Notably, there is a meaningful relation between the two activities. In fact, if the RF field, measured inside an MRI scanner, leads to acquire quantitative information about the model, the knowledge of the model is an essential factor for the shaping strategy to ensure a field which is as homogeneous as possible. Both proposed methods have several innovations respect to the state of the art and allow computational advantages compared to standard methods usually employed. Finally, both activities are validated through numerical experiments, tested against 2D simulated human brain phantom

Questa tesi si concentra sullo sviluppo di soluzioni avanzate per problemi di recupero e sintesi elettromagnetica rilevanti per la risonanza magnetica (MRI). Due problemi aperti nella risonanza magnetica sono: 1) lo sfruttamento di questa modalità di imaging per quantificare le proprietà fisiche, chimiche o fisiologiche utili per diverse applicazioni biomediche, ad es. trattamenti terapeutici e diagnosi; 2) la necessità di ottenere immagini ad alta risoluzione, minimizzando qualsiasi tipo di artefatto che potrebbe apparire a causa dell'elevato livello di campo impiegato durante una scansione MRI. Per quanto riguarda la prima questione, una possibilità è l'uso del campo di radiofrequenza (RF) raccolto all'interno dello scanner MRI per ricavare in modo non invasivo le proprietà elettriche in vivo dei tessuti biologici, ovvero la procedura che prende il nome di Tomografia delle proprietà elettriche basata su MRI (MRI-EPT). D'altra parte, un parametro essenziale che può migliorare la qualità dell'immagine è l'omogeneità del campo RF, la procedura utilizzata per attuare questo obiettivo è nota come livellamento del campo a radiofrequenza o RF shimming. I risultati presentati nella Tesi riguardano nuovi approcci basati sulla fisica sviluppati sia per MRI-EPT che per RF shimming. A questo proposito, presentiamo nuove metodologie basate sull'apprendimento per MRI-EPT e, d'altra parte, una nuova procedura di shimming RF che affronta il problema come fosse un problema di sagomatura del campo. A differenza della maggior parte degli approcci di apprendimento che adottano le reti neurali come “black box”, le strategie di learning proposte tengono conto della fisica del problema fornita dal calcolo del gradiente. Per quanto riguarda la procedura di shimming messa a punto, essa è stata ulteriormente ottimizzata con lo sviluppo di un modello ausiliario per la scelta conveniente del parametro da impiegare nella procedura di ottimizzazione. In particolare, esiste una relazione significativa tra le due attività. Infatti, se il campo RF, misurato all'interno di uno scanner MRI, porta ad acquisire informazioni quantitative sul modello, la conoscenza del modello è un fattore essenziale per la strategia di modellamento per garantire un campo il più omogeneo possibile. Entrambi i metodi proposti presentano numerose innovazioni rispetto allo stato dell'arte e consentono vantaggi computazionali rispetto ai metodi standard normalmente impiegati. Infine, entrambe le attività sono convalidate attraverso esperimenti numerici, testati nel caso di testa umana simulata in 2D

Electromagnetic recovery and synthesis problems in MRI: new approaches to Ept and Rf-shimming / Zumbo, Sabrina. - (2023 Apr 03).

Electromagnetic recovery and synthesis problems in MRI: new approaches to Ept and Rf-shimming

Zumbo, Sabrina
2023-04-03

Abstract

This Thesis focuses on the development of advanced solutions for electromagnetic recovery and synthesis problems relevant for Magnetic Resonance Imaging (MRI). Two open issues in MRI are: 1) the exploitation of this imaging modality to theoretically quantify physical, chemical, or physiological property to be useful for different biomedical application, f.i. therapeutic treatments and diagnosis; 2) the necessity to obtain high resolution images, minimizing any kind of artifacts that may appear due to the high field level employed during an MRI scan. As far as the first issue is concerned, a possibility is the use of the radiofrequency (RF) field collected inside the MRI scanner to non-invasively retrieve the in-vivo electrical properties of biological tissues, that is the so-called MRI-based Electrical Properties Tomography. On the other hand, an essential parameter that can improve to enhance the image quality is the homogeneity of the RF field, referred as RF shimming. Results presented in this Thesis concern new physics-assisted approaches developed for both MRI-EPT and RF shimming. In this respect, we present novel learning-based methodologies for MRI-EPT, and, on the other hand, a new RF shimming procedure that addresses the underlying issue as a field shaping problem. Unlike most of the learning approaches that adopt neural networks as “black-boxes”, the proposed learning-based strategies, take into account the physics of the problem provided by the computation of the gradient. As far as the shimming procedure, it is further optimized with the development of an auxiliary model for the convenient selection of the parameter to be employed in the shaping optimization. Notably, there is a meaningful relation between the two activities. In fact, if the RF field, measured inside an MRI scanner, leads to acquire quantitative information about the model, the knowledge of the model is an essential factor for the shaping strategy to ensure a field which is as homogeneous as possible. Both proposed methods have several innovations respect to the state of the art and allow computational advantages compared to standard methods usually employed. Finally, both activities are validated through numerical experiments, tested against 2D simulated human brain phantom
3-apr-2023
Settore ING-INF/02 - CAMPI ELETTROMAGNETICI
ISERNIA, Tommaso
IERA, Antonio
Doctoral Thesis
File in questo prodotto:
File Dimensione Formato  
Zumbo Sabrina.pdf

accesso aperto

Tipologia: Tesi di dottorato
Licenza: DRM non definito
Dimensione 11.89 MB
Formato Adobe PDF
11.89 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.12318/136766
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact