The magnetic susceptibility is a fundamental property of matter, which measures the degree of magnetization of a material when it is exposed to a magnetic field. Studying tissue susceptibility helps in diagnosis and treatment of disorders, especially in the brain. Quantitative Susceptibility Mapping (QSM) is an advanced MRI technique that quantifies tissue magnetic susceptibility, providing insights into tissue composition. QSM applies an inversion model to phase measurements returning the susceptibility distribution voxelwise even if the positive and negative susceptibility sources cannot be disentangled. The inversion is an ill-posed problem, thus implicating noise and artifacts in the reconstructed map. The analysis of QSM maps highlights their inability to distinguish between distinct susceptibility sources. QSM generates combined maps that represent a mixture of positive and negative contributions, hindering source separation. To overcome this problem, different strategies have been implemented to reconstruct QSM maps. This thesis presents a comprehensive analysis of the susceptibility separation technique proposed by Shin et al., and its application in comparison to conventional QSM. This susceptibility separation technique offers insights into brain microstructure, particularly the distribution of magnetic sources like iron and myelin. This is fundamental for diagnosis of neurological diseases because these sources are preferentially located in specific brain structures, such as Deep Grey Matter (DGM) for iron and axonal bundles for myelin. The study primarily aims to critically examine the assumption underlying the technique, the fixed relaxometric constant D_r,pos set at 137 Hz/ppm. This constant significantly impacts the characterization of magnetic properties within specific brain regions. Various computational approaches for estimating D_r,pos were explored, elucidating their influence on the separation of susceptibility sources. Furthermore, the study applies the susceptibility-separation algorithm on a cohort affected by Long COVID to unveil the potential of the technique to differentiate susceptibility sources within brain regions. The research is conducted on a dataset of 27 Long COVID subjects and 19 healthy controls. The study focuses on the potential neurological impact of Long COVID starting from the examination of QSM preprocessing, adhering to guidelines from the QSM Consensus Organization Committee. The study demonstrates that the order of QSM preprocessing steps, particularly background field removal and phase unwrapping, does not affect QSM maps. Then, the preprocessing pipeline is applied consistently to the entire cohort. The research critically evaluates the susceptibility-separation method proposed by Shin et al., that allows to obtain positive and negative susceptibility maps. A user-friendly graphical interface for clinicians has been developed and it is downloaded and explored in this work. A limitation of this technique is that D_r is equal for positive and negative contributions and it is fixed, thus it does not take into account any tissue composition variations and brain anatomy differences among subjects. This thesis delves into the segmentation of DGM regions of interest to assess the dependence of D_r,pos on specific DGM nuclei. D_r,pos distribution was assessed using histograms for different DGM nuclei both in healthy and Long COVID patients. Then, to investigate whether D_r,pos depends on the pathological state the relationship between R2‘ and conventional QSM values of DGM nuclei was evaluated with a linear regression analysis for the two groups.In conclusion, this thesis provides a comprehensive analysis of one susceptibility separation method effectiveness in characterizing brain microstructure. The study highlights the non-constant and subject-dependent nature of D_r, shedding light on magnetic properties in the brain, particularly in neurological conditions such as Long COVID.
La suscettività magnetica è una proprietà fondamentale della materia che misura il grado di magnetizzazione di un materiale quando è esposto a un campo magnetico. Lo studio della suscettività dei tessuti aiuta nella diagnosi e nel trattamento di disturbi, specialmente nel cervello. La Quantitative Susceptibility Mapping (QSM) è una tecnica avanzata di risonanza magnetica che quantifica la suscettività magnetica dei tessuti, dando informazioni sulla composizione dei tessuti. La QSM applica un modello di inversione alle misurazioni di fase per restituire la distribuzione di suscettività, anche se le sorgenti positive e negative non possono essere distinte. L'inversione è un problema mal posto, il che implica la presenza di rumore ed artefatti nella mappa. L'analisi delle mappe QSM evidenzia l'incapacità di distinguere tra diverse sorgenti di suscettività. La QSM genera mappe che rappresentano una miscela di contributi positivi e negativi, ostacolandone la separazione. Per superare questo problema, sono state implementate diverse strategie per la ricostruzione delle mappe QSM. Questa tesi presenta un'analisi completa del metodo di separazione della suscettività proposto da Shin et al. e la sua applicazione in confronto alla QSM. Questo metodo offre informazioni sulla microstruttura del cervello, in particolare sulla distribuzione di sorgenti magnetiche come il ferro e la mielina. Ciò è fondamentale per la diagnosi di malattie neurologiche perché queste sorgenti si trovano in specifiche strutture cerebrali, come la Sostanza Grigia Profonda (DGM) per il ferro e i fasci di fibre nervose per la mielina. Lo studio mira a esaminare criticamente l'assunzione alla base della tecnica, la costante rilassometrica D_r,pos fissata a 137 Hz/ppm. Questa costante influisce significativamente sulla caratterizzazione delle proprietà magnetiche in specifiche regioni cerebrali. Sono stati esplorati varie approcci computazionali per stimare D_r,pos. Inoltre, lo studio applica l'algoritmo di separazione della suscettività a un gruppo di soggetti Long COVID per valutare il potenziale della tecnica nel differenziare le sorgenti di suscettività nelle regioni cerebrali. La ricerca è stata condotta su un dataset di 27 soggetti Long COVID e 19 soggetti sani. Lo studio si concentra sul potenziale impatto neurologico del Long COVID a partire dall'analisi del QSM preprocessing, seguendo le linee guida del QSM Consensus Organization Committee. La ricerca dimostra che l'ordine delle fasi di QSM preprocessing, in particolare la rimozione del background e il phase unwrapping, non influisce sulle mappe QSM. Il QSM preprocessing viene applicato in modo coerente all'intero gruppo. La ricerca valuta il metodo di separazione della suscettività proposto da Shin et al., che consente di avere mappe di suscettività positive e negative. È stata sviluppata un'interfaccia grafica per i medici, esplorata in questo lavoro. Un limite è che D_r è uguale e fisso per i contributi positivi e negativi, quindi non tiene conto di variazioni nella composizione dei tessuti e differenze anatomiche tra i soggetti. Questa tesi approfondisce la segmentazione delle regioni di interesse per valutare la dipendenza di D_r,pos da specifici nuclei della DGM. La distribuzione di D_r,pos è stata valutata con istogrammi per diversi nuclei di DGM sia nei sani che nei Long COVID. Successivamente, per indagare se D_r,pos dipenda dallo stato patologico, è stata valutata la relazione tra i valori di R2‘ e di QSM nei nuclei della DGM con un'analisi di regressione lineare per i due gruppi. In conclusione, questa tesi fornisce un'analisi completa dell'efficacia del metodo di separazione della suscettività nella caratterizzazione della microstruttura cerebrale. Lo studio evidenzia la natura non costante e soggetto-dipendente D_r, gettando luce sulle proprietà magnetiche nel cervello, in particolare nelle condizioni neurologiche come il Long COVID.
Quantitative Susceptibility Mapping e Metodi di χ-Separation: come la RM può aiutarci a comprendere il Long COVID
MEGALIZZI, SILVIA
2022/2023
Abstract
The magnetic susceptibility is a fundamental property of matter, which measures the degree of magnetization of a material when it is exposed to a magnetic field. Studying tissue susceptibility helps in diagnosis and treatment of disorders, especially in the brain. Quantitative Susceptibility Mapping (QSM) is an advanced MRI technique that quantifies tissue magnetic susceptibility, providing insights into tissue composition. QSM applies an inversion model to phase measurements returning the susceptibility distribution voxelwise even if the positive and negative susceptibility sources cannot be disentangled. The inversion is an ill-posed problem, thus implicating noise and artifacts in the reconstructed map. The analysis of QSM maps highlights their inability to distinguish between distinct susceptibility sources. QSM generates combined maps that represent a mixture of positive and negative contributions, hindering source separation. To overcome this problem, different strategies have been implemented to reconstruct QSM maps. This thesis presents a comprehensive analysis of the susceptibility separation technique proposed by Shin et al., and its application in comparison to conventional QSM. This susceptibility separation technique offers insights into brain microstructure, particularly the distribution of magnetic sources like iron and myelin. This is fundamental for diagnosis of neurological diseases because these sources are preferentially located in specific brain structures, such as Deep Grey Matter (DGM) for iron and axonal bundles for myelin. The study primarily aims to critically examine the assumption underlying the technique, the fixed relaxometric constant D_r,pos set at 137 Hz/ppm. This constant significantly impacts the characterization of magnetic properties within specific brain regions. Various computational approaches for estimating D_r,pos were explored, elucidating their influence on the separation of susceptibility sources. Furthermore, the study applies the susceptibility-separation algorithm on a cohort affected by Long COVID to unveil the potential of the technique to differentiate susceptibility sources within brain regions. The research is conducted on a dataset of 27 Long COVID subjects and 19 healthy controls. The study focuses on the potential neurological impact of Long COVID starting from the examination of QSM preprocessing, adhering to guidelines from the QSM Consensus Organization Committee. The study demonstrates that the order of QSM preprocessing steps, particularly background field removal and phase unwrapping, does not affect QSM maps. Then, the preprocessing pipeline is applied consistently to the entire cohort. The research critically evaluates the susceptibility-separation method proposed by Shin et al., that allows to obtain positive and negative susceptibility maps. A user-friendly graphical interface for clinicians has been developed and it is downloaded and explored in this work. A limitation of this technique is that D_r is equal for positive and negative contributions and it is fixed, thus it does not take into account any tissue composition variations and brain anatomy differences among subjects. This thesis delves into the segmentation of DGM regions of interest to assess the dependence of D_r,pos on specific DGM nuclei. D_r,pos distribution was assessed using histograms for different DGM nuclei both in healthy and Long COVID patients. Then, to investigate whether D_r,pos depends on the pathological state the relationship between R2‘ and conventional QSM values of DGM nuclei was evaluated with a linear regression analysis for the two groups.In conclusion, this thesis provides a comprehensive analysis of one susceptibility separation method effectiveness in characterizing brain microstructure. The study highlights the non-constant and subject-dependent nature of D_r, shedding light on magnetic properties in the brain, particularly in neurological conditions such as Long COVID.È consentito all'utente scaricare e condividere i documenti disponibili a testo pieno in UNITESI UNIPV nel rispetto della licenza Creative Commons del tipo CC BY NC ND.
Per maggiori informazioni e per verifiche sull'eventuale disponibilità del file scrivere a: unitesi@unipv.it.
https://hdl.handle.net/20.500.14239/16543