Diffusion-based Magnetic Resonance Imaging (dMRI) is rapidly becoming the instrument of choice to probe the structure of the human brain in vivo. By modelling the properties of water diffusion inside cerebral tissues, it is indeed possible to extract surrogates of histological measures, such as fibre density, conformation and preferential direction, in a non-invasive manner. Furthermore, local orientational features can be used to reconstruct axonal pathways that link different brain regions, allowing the study of how they are structurally connected. Nevertheless, the quantification of dMRI measures must be cautious when the physiological environment of brain tissues is drastically altered. Such is the case of brain tumours. The microstructure of brain tumours is highly heterogeneous, being diverse between and inside specific types and malignancy grade. The wide spectrum of cellular environments they feature invalidates several hypotheses on which diffusion-based microstructure models are built and, contemporarily, poses difficulties in the process of tracking white matter in affected regions. Given these limitations, are these techniques worth using in this complex pathological environment? During the last three years I explored several state of the art diffusion-based methodologies in a cohort of patients suffering from a range of brain tumours. Hence, this thesis strives to be a summary of this work, laying the foundation for future studies aiming to integrate the use of advanced dMRI in the clinical neuro-oncological practice. The thesis is divided in three main parts, which are organized as follows: In the first part, an assessment is made whether two widely known diffusion advanced models, Neurite Orientation Dispersion and Density Imaging (NODDI) and the Spherical Mean Technique (SMT) are properly fitted in the tumoral lesion in terms of goodness-of-fit and parameter precision. Several works, concentrating mainly on NODDI, used such techniques not as biophysical models but as signal representations, trying to find biomarkers that differentiate more and less isotropic environments which contribute to the totality of the diffusion signal in ‘tumoral’ voxels. These studies were performed without first checking whether these diffusion metrics are mathematically reliable. This issue is here assessed from a technical point of view, without giving specific biophysical meaning to the models in exam inside the tumoral tissues The second part features a comparison study between methods for the identification of structurally disconnected white matter (WM) in brain tumour patients. Here, two branches of methodologies were identified, namely direct and indirect approaches. The formers use single-subject tractography to directly investigate which fibre bundles may be affected by the presence of the tumour. The latters, instead, embed the focal lesion on a normative atlas of white matter tracts, identifying the probability of a WM voxel being disconnected by the pathology. Employing known image analysis metrics, both approaches are discussed, highlighting points of convergence, but also of disagreement, in terms of the physio-pathological information they can convey. In the third and last part of this thesis, tumour-related anomalies of diffusion-based structural connectivity (SC) matrices are put in relationship with metabolic measures from [18F]-FDG PET. A procedure for tractography algorithm selection was firstly performed, and after the SC quantification, a statistical method of detecting altered connections in the tumour-affected SC matrix is presented. Within such a framework, the amount of affected SC entries was eventually quantified in the available cohort of patients and put in relationship with standardized uptake values from PET. Finally, a discussion of the results of this association is provided, paying particular attention to the limitations of these imaging modalities in the brain oncological field.
La Risonanza magnetica di diffusione (dMRI) sta diventando lo strumento più adatto per indagare la microstruttura del cervello umano in vivo. Modellando le proprietà della diffusione dell’acqua nei tessuti cerebrali, è infatti possibile ottenere delle misure simili a quelle derivate dall’istologia, come la densità di fibre, la loro conformazione e la loro direzione di propagazione, in maniera non invasiva. In più, misure locali di integrità e di orientazione della materia bianca possono essere usate da algoritmi di trattografia per ricostruire globalmente il percorso seguito dalle fibre in tutto il cervello, permettendo di studiare come le varie regioni corticali sono connesse. Nonostante ciò, l’utilizzo della dMRI deve essere condotto con attenzione in presenza di patologie che alterano drasticamente la fisiologia del cervello, come nel caso dei tumori cerebrali. La varietà di microambienti cellulari che caratterizza questo tipo di patologie invalida alcune ipotesi sul quale si fondano i modelli di microstruttura basati sulla dMRI. In più, il processo di ricostruzione della trattografia nel cervello presenta particolari difficoltà tecniche nelle regioni affette dalla patologia. Date queste limitazioni, vi è del valore nell’utilizzare tecniche basate sulla dMRI in questo complesso ambiente patologico? Negli ultimi tre anni, ho avuto modo di esplorare diverse di queste metodologie in una popolazione di pazienti con tumore cerebrale. La presente tesi vorrebbe quindi essere una sintesi di questo lavoro, che costituisce una base verso l’integrazione di tecniche di diffusione avanzate all’interno della pratica neuro-oncologica. Nella sua interità, la tesi presenta tre lavori, organizzati come segue: La prima parte presenta uno studio analitico su due noti modelli di microstruttura, Neurite Orientation Dispersion and Density Imaging (NODDI) e la Spherical Mean Technique. Questo lavoro è volto alla quantificazione della bontà del fit e precisione parametrica delle due tecniche all’interno della lesione tumorale. Alcuni lavori, concentrati principalmente su NODDI, usano queste tecniche come modelli di segnale e non biofisici, cercando di trovare biomarker capaci di caratterizzare aspecificamente il tessuto patologico. L’analisi qui svolta supporta i risultati di letteratura da un punto di vista tecnico, senza considerazioni sul significato biologico di questi modelli. La seconda parte contiene uno studio di confronto tra due diversi metodi per la quantificazione di regioni di materia bianca sconnessa a causa del tumore. Due categorie di approcci qui sono stati studiati: approcci diretti, e approcci indiretti. I primi fanno uso della trattografia singolo-soggetto per investigare quali fasci di fibre siano affetti nel loro decorso dalla presenza del tumore. I secondi invece non hanno bisogno di acquisizioni dMRI, e utilizzano un atlante normativo di fasci di materia bianca per investigare, probabilisticamente, quali di questi potrebbero essere affetti data la locazione e l'estensione della zona tumorale. Utilizzando noti strumenti di analisi dell’immagine, i due approcci vengono qui confrontati, discutendo pregi e difetti di ciascun metodo. Nella terza e ultima parte della tesi, viene studiata la relazione tra alterazioni di matrici di connettività strutturale (SC) di pazienti tumorali e variazioni regionali di metabolismo misurate usando la Tomografia ad Emissione di Positroni (PET) con tracciante [18F]-FDG. All'interno di questo studio, viene prima proposta una procedura per la selezione dell’algoritmo di trattografia ottimale per le analisi. A seguire, viene sviluppata una metodologia statistica per rilevare le loro connessioni della matrice SC alterate dalla presenza del tumore. La presenza di queste alterazioni viene infine correlata con la PET, e si discutono i risultati ottenuti, ponendo particolare attenzione alle limitazioni di entrambe queste modalità di imaging.
Dall'imaging di microstruttura alla connettività strutturale: l'utilizzo della risonanza magnetica di diffusione per investigare l'impatto dei gliomi sul cervello umano / Villani, Umberto. - (2022 Apr 22).
Dall'imaging di microstruttura alla connettività strutturale: l'utilizzo della risonanza magnetica di diffusione per investigare l'impatto dei gliomi sul cervello umano
VILLANI, UMBERTO
2022
Abstract
Diffusion-based Magnetic Resonance Imaging (dMRI) is rapidly becoming the instrument of choice to probe the structure of the human brain in vivo. By modelling the properties of water diffusion inside cerebral tissues, it is indeed possible to extract surrogates of histological measures, such as fibre density, conformation and preferential direction, in a non-invasive manner. Furthermore, local orientational features can be used to reconstruct axonal pathways that link different brain regions, allowing the study of how they are structurally connected. Nevertheless, the quantification of dMRI measures must be cautious when the physiological environment of brain tissues is drastically altered. Such is the case of brain tumours. The microstructure of brain tumours is highly heterogeneous, being diverse between and inside specific types and malignancy grade. The wide spectrum of cellular environments they feature invalidates several hypotheses on which diffusion-based microstructure models are built and, contemporarily, poses difficulties in the process of tracking white matter in affected regions. Given these limitations, are these techniques worth using in this complex pathological environment? During the last three years I explored several state of the art diffusion-based methodologies in a cohort of patients suffering from a range of brain tumours. Hence, this thesis strives to be a summary of this work, laying the foundation for future studies aiming to integrate the use of advanced dMRI in the clinical neuro-oncological practice. The thesis is divided in three main parts, which are organized as follows: In the first part, an assessment is made whether two widely known diffusion advanced models, Neurite Orientation Dispersion and Density Imaging (NODDI) and the Spherical Mean Technique (SMT) are properly fitted in the tumoral lesion in terms of goodness-of-fit and parameter precision. Several works, concentrating mainly on NODDI, used such techniques not as biophysical models but as signal representations, trying to find biomarkers that differentiate more and less isotropic environments which contribute to the totality of the diffusion signal in ‘tumoral’ voxels. These studies were performed without first checking whether these diffusion metrics are mathematically reliable. This issue is here assessed from a technical point of view, without giving specific biophysical meaning to the models in exam inside the tumoral tissues The second part features a comparison study between methods for the identification of structurally disconnected white matter (WM) in brain tumour patients. Here, two branches of methodologies were identified, namely direct and indirect approaches. The formers use single-subject tractography to directly investigate which fibre bundles may be affected by the presence of the tumour. The latters, instead, embed the focal lesion on a normative atlas of white matter tracts, identifying the probability of a WM voxel being disconnected by the pathology. Employing known image analysis metrics, both approaches are discussed, highlighting points of convergence, but also of disagreement, in terms of the physio-pathological information they can convey. In the third and last part of this thesis, tumour-related anomalies of diffusion-based structural connectivity (SC) matrices are put in relationship with metabolic measures from [18F]-FDG PET. A procedure for tractography algorithm selection was firstly performed, and after the SC quantification, a statistical method of detecting altered connections in the tumour-affected SC matrix is presented. Within such a framework, the amount of affected SC entries was eventually quantified in the available cohort of patients and put in relationship with standardized uptake values from PET. Finally, a discussion of the results of this association is provided, paying particular attention to the limitations of these imaging modalities in the brain oncological field.File | Dimensione | Formato | |
---|---|---|---|
Tesi_Umberto_Villani.pdf
accesso aperto
Descrizione: Tesi_Umberto_Villani
Tipologia:
Tesi di dottorato
Dimensione
24.85 MB
Formato
Adobe PDF
|
24.85 MB | Adobe PDF | Visualizza/Apri |
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.