Magnetic Resonance Imaging (MRI) was firstly introduced between the ’70 and ’80 years. Since its introduction, there has been a rapid diffusion of the MRI technique, because of the good quality of the images and the enlarged fields of application both for clinical purposes and for research. Recent technological advancements have allowed the diffusion of MRI scanners with ultra-high magnetic field strength. Structural and functional MRI at ultra-high field have allowed the study of cerebral anatomy and function with an increased spatial resolution and tissue contrast. In this thesis, we analysed functional and structural MRI images of the brain acquired at ultra-high field (7T) from 10 couples of twins (6 MZ, 4 DZ, 4 discordant and 6 not discordant for risk of psychosis), with the final aim to identify cerebral correlates of psychotic symptoms and to distinguish the effects of environmental and genetical factors. Due to the intensity inhomogeneities of the structural images at 7T, the automatic pre-processing provided by the most commonly used software tools for biomedical image processing (SPM, Freesurfer, and FSL) is heavily affected by artefacts and produces non satisfactory results. For this reason, we compared different algorithms and parameters for bias magnetic field regularisation and cerebral tissue segmentation in terms of results quality, in order to reduce intensity inhomogeneity and adequately identify brain tissues, and in turn get more reliable results. We selected the best algorithm of bias correction and the optimal parameters by comparing the results through an accurate visual inspection. Based on the obtained results, we found that the 7T structural images with the most effective artefact correction come from the algorithm applied by SPM 12, therefore we selected this algorithm for the subsequent analyses. Furthermore, for a quantitative evaluation of the quality of artefact removal in the results, we performed a statistical comparison among the 7T images before and after bias correction and the 3T images recorded from the same subjects, using quantitative morphological parameters extracted from Freesurfer (i.e. cortical thickness, cortical surface area, mean cortical curvature, white matter and grey matter cortical and subcortical volumes; these parameters are extracted for regions defined according to a specific anatomical atlas). For each morphological parameter, we performed two series of paired t-tests: (1) between images at 7T before-correction and images at 3T; (2) between t images at 7T after-correction and images at 3T. Based on visual inspection of the results quality, the significant differences emerged between 3T images and 7T original images were associated with noise contribution. After SPM12 bias correction of the 7T images, the obtained results were more reliable and comparable to the ones obtained at 3T, which lets us assume that the lower parameter distance between 7T after-correction and 3T can be attributed to the efficient artefact removal. The residual difference between images at 7T after-correction and images at 3T may be due to the better quality of images at 7T (better resolution, better signal to noise ratio and contrast to noise ratio, and partial volume effect reduction) that may produce more realistic estimates. One of the main purposes of the research was the identification of regions of the brain with functional and structural abnormalities due to the risk of psychosis. For the identification of regions with structural correlates of psychotic risk, we used the structural images at 7T from which we extracted white matter and grey matter volumes for all the anatomical regions defined according to the AAL atlas (Automated Anatomical Labeling, for grey matter) and JHU-White Matter Labels (for white matter). The fMRI (functional Magnetic Resonance Imaging) study in resting state was aimed to identify functional connectivity alterations correlated to the risk of psychosis: we performed a seed-based analysis in the subjects’ native space (including only those regions generally involved in psychotic disorders according to the literature), which produced the adjacency matrices from which we extracted the network topological parameters. For the interpretation of the results from the anatomical and functional analyses, we assumed the existence of three different levels of risk of psychosis on the population: (1) no risk, for the control couples of twins; (2) low risk, in the twin of subjects with psychotic risk, due to the sharing of genetical and environmental factors; (3) high risk, in the subjects with over-threshold psychotic risk identified with the rating scale ERIraos-CL. In order to identify cerebral correlates of psychotic symptoms, we performed the following statistical analysis on the previously extracted parameters (grey matter and white matter volumes, connection values and topological parameters): (1) parametric (or nonparametric) paired test between high risk subjects and their twins (low-risk subjects); (2) parametric (or nonparametric) two sample test between high risk twins and healthy subjects (single individuals from healthy twin couples); (3) parametric (or nonparametric) sample test between low risk twins and healthy subjects. Furthermore, to verify if discordant twins are more/less similar between each other than non-discordant twins, we applied general linear model analyses on the absolute values of intra-couple difference for the previously extracted parameters. We assumed that regions with functional and anatomical abnormalities due to a high risk of psychosis would emerge from the comparisons (p-value<0.05) between high-risk subjects and their siblings as well as healthy subjects. In these regions, discordant twins should be significantly (p-value<0.01) less similar than non-discordant twins according to the GLM. Conversely, regions altered due to a low risk of psychosis are those get both from the comparison (p-value<0.05) between healthy subjects and both high-risk and low-risk subjects and regions in which discordant twins are significantly (p-value<0.01) more similar between each other than non-discordant twins according to the GLM. The results allowed us to identify the hippocampus, the inferior frontal middle gyrus, and the left insula, as regions with anatomical and functional abnormalities due to a high risk, and some frontal regions (the “Frontal Inf Oper R” e “Frontal Inf Orb R”) as regions with anatomical and functional abnormalities due to a low risk of psychosis. These results could represent neurobiological markers for the risk of psychotic disorders in young people, allowing the fast recognition of subjects at risk and the development of innovative therapeutic strategies.
Sin dalla sua introduzione in clinica a partire dagli anni ’70 e ‘80, l’MRI ha avuto una grande diffusione e un grande sviluppo, sia per la qualità delle immagini acquisite, sia per la versatilità di applicazione in ambito clinico e di ricerca. L’incremento dell’intensità del campo magnetico ha permesso di migliorare lo studio sia della struttura che della funzione del cervello, grazie all’aumentato rapporto segnale rumore e contrasto rumore e alla migliore risoluzione delle immagini. Quindi l’impego di campi magnetici ad elevata intensità ha lo scopo di ottenere informazioni su struttura e funzionalità cerebrale in vivo ad un livello di dettaglio senza precedenti. Nella presente tesi sono stati sviluppati algoritmi per l’analisi di immagini neuroanatomiche e neurofunzionali acquisite a campo ultra-alto (7T) da 10 giovani coppie di gemelli, di cui 4 dizigoti e 6 monozigoti, 4 discordanti e 6 non discordanti per rischio psicotico, con l’obiettivo ultimo di identificare le basi cerebrali del rischio di sviluppare psicosi e separare l’effetto di fattori genetici e ambientali. Poiché le immagini neuroanatomiche acquisite a campo ultra-alto sono comunemente affette da disomogeneità di intensità che ne rendono problematica l’elaborazione automatica, sono stati ottimizzati i parametri di correzione degli artefatti e confrontate le performance dei principali software di elaborazione di neuroimmagini (SPM, Freesurfer, e FSL), al fine di ottimizzare i risultati delle analisi successive. A tale scopo, sono stati confrontati diversi parametri di regolarizzazione del bias del campo magnetico e di segmentazione dei tessuti cerebrali, al fine di adattare al meglio le elaborazioni sulle immagini ad alta risoluzione basandosi sull’accurata ispezione visiva dei risultati. Tra gli algoritmi di correzione relativi ai diversi software, è stato selezionato per le analisi successive l’algoritmo eseguito da SPM 12, che conduceva alla correzione più soddisfacente. Inoltre, per una valutazione quantitativa degli effetti dei passaggi di correzione degli artefatti, sono stati confrontati parametri quantitativi relativi alle caratteristiche morfologiche del cervello (quali spessore, area superficiale, curvatura media e volumi regionali di materia grigia calcolati per regioni anatomiche definite secondo opportuni atlanti) ottenuti dalle analisi di Freesurfer sa partire dalle immagini a 7T prima e dopo correzione delle disomogeneità con SPM 12 e dalle immagini a 3T acquisite dagli stessi soggetti. In particolare, sono stati eseguite due serie di t-test accoppiati, rispettivamente tra 1) i valori dei parametri calcolati dalle immagini a 7T originali e dalle rispettive immagini a 3T, e 2) tra i valori dei parametri dalle immagini a 7T post-correzione e dalle rispettive immagini a 3T. Dal momento che, a seguito della correzione delle disomogeneità di SPM, i risultati delle immagini a 7T risultano più simili a quelli ottenuti dalle immagini a 3T rispetto ai risultati ottenuti dalle immagini a 7T prima della correzione, è ragionevole attribuire questa minore differenza all’eliminazione efficace degli artefatti. Inoltre, si può ipotizzare che le differenze residue tra le immagini a 7T corrette e le immagini a 3T siano dovute al maggiore livello di dettaglio e al risaltato contrasto tessutale delle immagini ad alta risoluzione, che consentirebbero stime più realistiche. Uno degli scopi principali della ricerca ha riguardato l’identificazione delle regioni del cervello caratterizzate da alterazioni anatomico-funzionali dovute al rischio di manifestare psicosi. Ricorrendo alle immagini strutturali a 7T, al fine di individuare le regioni affette da alterazioni anatomiche, si sono estratti i volumi di materia grigia e bianca per ognuna di esse (definite secondo gli atlanti AAL per le analisi sui volumi di materia grigia, e JHU per materia bianca). Lo studio di neuroimaging funzionale, sulle immagini fMRI (functional Magnetic Resonance Imaging) a 7T in resting state, è stato finalizzato a individuare alterazioni di connettività funzionale associabili al rischio psicotico. A tale scopo si è applicata l’analisi di connettività seed-based in spazio nativo (selezionando le regioni comunemente coinvolte nei disturbi psicotici) che ha permesso di calcolare le matrici di connettività per ogni soggetto, da cui sono stati estratti i parametri topologici delle reti. Per l’interpretazione dei risultati, si è ipotizzata la presenza di tre livelli di rischio presenti nella popolazione: (1) rischio zero, per i gemelli di controllo (i non discordanti), (2) basso rischio, in comune ai gemelli discordanti, dovuto alla condivisione di fattori genetici e ambientali, (3) alto rischio, per uno dei due gemelli discordanti (identificato mediante opportune scale di valutazione). Al fine di identificare le basi cerebrali del rischio di manifestare psicosi nella popolazione di gemelli, sono state condotte le seguenti analisi statistiche su tutti i parametri estratti (volumi, valori di connettività tra le regioni, valori di connettività nodale, e parametri topologici): (1) test parametrico (o non parametrico) accoppiato tra i gemelli ad alto rischio e i rispettivi gemelli a basso rischio; (2) test parametrico (o non parametrico) tra i gemelli ad alto rischio e i gemelli di controllo; (3) test parametrico (o non parametrico) tra i gemelli a basso rischio e i gemelli di controllo. Inoltre, per verificare se i gemelli discordanti siano più simili (o dissimili) rispetto ai gemelli non discordanti, si è applicato il modello lineare generale sui valori assoluti delle differenze intracoppia di tutti i gemelli, per i parametri precedentemente annoverati. Si è ipotizzato che le regioni coinvolte nell’alto rischio di manifestare psicosi, sono quelle ottenute dal test accoppiato (p-value<0,05) tra gemelli ad alto rischio vs basso rischio, e quelle regioni in cui i gemelli discordanti sono significativamente più dissimili (p-value<0,01) rispetto ai gemelli non discordanti; le regioni aventi alterazioni anatomico-funzionali a causa di un basso rischio di manifestare psicosi, sono quelle ottenute sia dal test basso rischio vs controllo, sia dal test alto rischio vs controllo, e le regioni in cui secondo il modello lineare generale i gemelli discordanti sono significativamente (p-value<0,01) più simili rispetto ai gemelli non discordanti. I risultati ottenuti dai test appena descritti hanno permesso di individuare 1) l’ippocampo, il giro frontale mediale sinistro, e l’insula sinistra come regioni caratterizzate da alterazioni anatomico-funzionali in presenza di alto rischio di manifestare psicosi; 2) il giro frontale opercolare destro inferiore e il giro orbito frontale destro inferiore come caratterizzati da alterazioni anatomico-funzionali legate a fattori di rischio genetici, comuni alle coppie di gemelli di cui almeno uno manifesta il rischio. I risultati ottenuti potrebbero rappresentare marker neurobiologici per il rischio di sviluppare psicosi, permettendo il riconoscimento precoce di soggetti a rischio e lo sviluppo di strategie terapeutiche innovative.
Studio delle basi cerebrali del rischio psicotico tramite imaging a risonanza magnetica a campo ultra-alto su giovani coppie di gemelli
TRICOLI, MIRIANA
2017/2018
Abstract
Magnetic Resonance Imaging (MRI) was firstly introduced between the ’70 and ’80 years. Since its introduction, there has been a rapid diffusion of the MRI technique, because of the good quality of the images and the enlarged fields of application both for clinical purposes and for research. Recent technological advancements have allowed the diffusion of MRI scanners with ultra-high magnetic field strength. Structural and functional MRI at ultra-high field have allowed the study of cerebral anatomy and function with an increased spatial resolution and tissue contrast. In this thesis, we analysed functional and structural MRI images of the brain acquired at ultra-high field (7T) from 10 couples of twins (6 MZ, 4 DZ, 4 discordant and 6 not discordant for risk of psychosis), with the final aim to identify cerebral correlates of psychotic symptoms and to distinguish the effects of environmental and genetical factors. Due to the intensity inhomogeneities of the structural images at 7T, the automatic pre-processing provided by the most commonly used software tools for biomedical image processing (SPM, Freesurfer, and FSL) is heavily affected by artefacts and produces non satisfactory results. For this reason, we compared different algorithms and parameters for bias magnetic field regularisation and cerebral tissue segmentation in terms of results quality, in order to reduce intensity inhomogeneity and adequately identify brain tissues, and in turn get more reliable results. We selected the best algorithm of bias correction and the optimal parameters by comparing the results through an accurate visual inspection. Based on the obtained results, we found that the 7T structural images with the most effective artefact correction come from the algorithm applied by SPM 12, therefore we selected this algorithm for the subsequent analyses. Furthermore, for a quantitative evaluation of the quality of artefact removal in the results, we performed a statistical comparison among the 7T images before and after bias correction and the 3T images recorded from the same subjects, using quantitative morphological parameters extracted from Freesurfer (i.e. cortical thickness, cortical surface area, mean cortical curvature, white matter and grey matter cortical and subcortical volumes; these parameters are extracted for regions defined according to a specific anatomical atlas). For each morphological parameter, we performed two series of paired t-tests: (1) between images at 7T before-correction and images at 3T; (2) between t images at 7T after-correction and images at 3T. Based on visual inspection of the results quality, the significant differences emerged between 3T images and 7T original images were associated with noise contribution. After SPM12 bias correction of the 7T images, the obtained results were more reliable and comparable to the ones obtained at 3T, which lets us assume that the lower parameter distance between 7T after-correction and 3T can be attributed to the efficient artefact removal. The residual difference between images at 7T after-correction and images at 3T may be due to the better quality of images at 7T (better resolution, better signal to noise ratio and contrast to noise ratio, and partial volume effect reduction) that may produce more realistic estimates. One of the main purposes of the research was the identification of regions of the brain with functional and structural abnormalities due to the risk of psychosis. For the identification of regions with structural correlates of psychotic risk, we used the structural images at 7T from which we extracted white matter and grey matter volumes for all the anatomical regions defined according to the AAL atlas (Automated Anatomical Labeling, for grey matter) and JHU-White Matter Labels (for white matter). The fMRI (functional Magnetic Resonance Imaging) study in resting state was aimed to identify functional connectivity alterations correlated to the risk of psychosis: we performed a seed-based analysis in the subjects’ native space (including only those regions generally involved in psychotic disorders according to the literature), which produced the adjacency matrices from which we extracted the network topological parameters. For the interpretation of the results from the anatomical and functional analyses, we assumed the existence of three different levels of risk of psychosis on the population: (1) no risk, for the control couples of twins; (2) low risk, in the twin of subjects with psychotic risk, due to the sharing of genetical and environmental factors; (3) high risk, in the subjects with over-threshold psychotic risk identified with the rating scale ERIraos-CL. In order to identify cerebral correlates of psychotic symptoms, we performed the following statistical analysis on the previously extracted parameters (grey matter and white matter volumes, connection values and topological parameters): (1) parametric (or nonparametric) paired test between high risk subjects and their twins (low-risk subjects); (2) parametric (or nonparametric) two sample test between high risk twins and healthy subjects (single individuals from healthy twin couples); (3) parametric (or nonparametric) sample test between low risk twins and healthy subjects. Furthermore, to verify if discordant twins are more/less similar between each other than non-discordant twins, we applied general linear model analyses on the absolute values of intra-couple difference for the previously extracted parameters. We assumed that regions with functional and anatomical abnormalities due to a high risk of psychosis would emerge from the comparisons (p-value<0.05) between high-risk subjects and their siblings as well as healthy subjects. In these regions, discordant twins should be significantly (p-value<0.01) less similar than non-discordant twins according to the GLM. Conversely, regions altered due to a low risk of psychosis are those get both from the comparison (p-value<0.05) between healthy subjects and both high-risk and low-risk subjects and regions in which discordant twins are significantly (p-value<0.01) more similar between each other than non-discordant twins according to the GLM. The results allowed us to identify the hippocampus, the inferior frontal middle gyrus, and the left insula, as regions with anatomical and functional abnormalities due to a high risk, and some frontal regions (the “Frontal Inf Oper R” e “Frontal Inf Orb R”) as regions with anatomical and functional abnormalities due to a low risk of psychosis. These results could represent neurobiological markers for the risk of psychotic disorders in young people, allowing the fast recognition of subjects at risk and the development of innovative therapeutic strategies.File | Dimensione | Formato | |
---|---|---|---|
Tesi Miriana Tricoli.pdf
accessibile in internet per tutti
Descrizione: Tesi di laurea magistrale di Miriana Tricoli
Dimensione
4.81 MB
Formato
Adobe PDF
|
4.81 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/146123