Background and Aim A comprehensive understanding of the neural mechanisms driving brain plasticity is pivotal for implementing tailored approaches promoting recovery from neurological conditions and restoring cognitive functions. Despite being the gold standard for mapping brain function, the use of fMRI in the longitudinal framework is currently characterized by several challenges ranging from the restrictive acquisition setting to the variability affecting the measures. In this context, this PhD project, developed at the CADiTeR Laboratory of IRCCS Fondazione Don Carlo Gnocchi in Milan, addresses some of these challenges to pave the way for the use of fMRI in the clinical rehabilitative setting, potentially leading to the identification of robust imaging derived biomarkers of restored brain function. Specifically, this project encompasses two major aims: 1) the integration of fMRI with fNIRS modality, offering a less restrictive and more ecological acquisition setting particularly suitable for longitudinal evaluation and multiple measurements; 2) the implementation of careful fMRI paradigm and analysis methods to take into account the evaluation of both intra- and inter-subject variability characterizing fMRI measurements. Methods fMRI-fNIRS Integration and Reliability Assessment – Chapter 2 Eighteen healthy volunteers underwent separate fMRI and fNIRS acquisition performing the same motor (hand grasping) paradigm, alternating right and left hand movements. The functional MRI data were preprocessed using FreeSurfer FS-FAST and were successively imported in Brainstorm, using user-defined scripts to perform integration with fNIRS data. The fNIRS dataset was preprocessed using standardized pipeline implemented in Nirstorm package of Brainstorm. Anatomicallly constrained functional ROIs were derived by thresholding fNIRS and fMRI statistical maps (punc <0.05, punc <0.001, pFDR <0.05) within bilateral motor parcels (pre and post central gyri) segmented according to the Desikan-Killany atlas. Spatial and temporal correspondence between the two techniques was assessed computing Dice Coefficient (DC) and Pearson’s Correlation respectively. Reliability and Reproducibility of an Emotion Generation Task fMRI Paradigm – Chapter 3 Sixty-two healthy subjects were enrolled in the study and performed two fMRI acquisition during the acquisition of an emotion stimulation fMRI task paradigm implemented in two forms including different but equivalent visual stimuli. fMRI data were preprocessed using a standard pipeline in SPM-12. Two-different contrasts were derived. Test-retest reliability and reproducibility (i.e. the equivalence of the parallel forms) of fMRI responses and patterns were evaluated. The analyses were performed voxel-wise, using the general linear model (GLM), and via a region-of-interest (ROI)-based approach, by computing the intra-class correlation coefficient (ICC) on the obtained contrasts. Assessment of inter-subject variability of an emotion stimulation task: application on a cohort of MS patients – Chapter 4 Thirty healthy volunteers and thirteen patients with multiple sclerosis (pwMS), showing depression symptoms as MS-associated comorbidity, were enrolled. pwMS were treated according to an eye-movement desensitization and reprocessing (EMDR) rehabilitation intervention. The healthy control group underwent MRI examination during two separate acquisitions, in which two different parallel forms of the same emotion stimulation task were delivered, mirroring the experimental settings adopted for the pwMS, which were scanned twice before and after rehabilitation. fMRI data were preprocessed using a standard pipeline in SPM-12 and standard GLM group statistic. Furthermore, the normalized T-maps were derived for each contrast at the individual level and used to derive threshold-weighted overlap map (〖OM〗_(thr-w)) assessing inter-subject variability. Both GLM activation maps and OM were masked using a user-defined inclusive mask comprising 16 bilateral ROIs salient for the emotional task. Assessment of inter-subject variability of an ecological social cognition task – Chapter 5 Thirty-five healthy controls underwent an fMRI acquisition during the administration of an adaptation of the ESCoT behavioural test. The task was implemented considering two conditions: the experimental Theory of Mind (ToM) condition versus the control physical inference (PI) condition, each including implicit mentalizing (movie watching), and explicit mentalizing (questions answering). SPM 12 was used for fMRI preprocessing following a standard pipeline. Seven contrasts comparing ToM (implicit/explicit) vs PI reasoning were derived. At group level standard GLM statistics were derived and further complemented by deriving overlap-maps to evaluate inter-subject variability. Both these analyses were performed considering a theory-driven approach using an atlas-derived inclusive mask depicting the Abu-Akel & Shamay-Tsoory model. Results fMRI-fNIRS Integration and Reliability Assessment – Chapter 2 The new implemented integration approach allowed a direct vertex-wise comparison between fMRI and fNIRS both considering a spatial and temporal correspondence assessment. At the group level the spatial correspondence, measured using DC, spanned from substantial to slight agreement moving from punc< 0.05 as the least restrictive to punc< 0.001 as the most conservative threshold. Similar results were also observed at individual level. A moderate to good correlation was observed overall at the group level concerning the temporal comparison. Reliability and Reproducibility of an Emotion Generation Task fMRI Paradigm – Chapter 3 No significant differences for both reliability and reproducibility analyses were revealed using voxel-wise GLM analysis neither for positive nor negative contrast. Considering instead the ROI-based approach, the test-retest reliability was poor (ICC<0.4) for the positive contrast and ranged from poor to excellent (0.4≤ICC≤0.75) for the negative one. Similar results were observed comparing the two different parallel forms (reproducibility analysis) with overall higher fMRI signal stability for negative stimuli over positive ones. Assessment of inter-subject variability of an emotion stimulation task: application on a cohort of MS patients – Chapter 4 No significant differences (pFWE< 0.05) were detected when comparing the 1st and 2nd scans, neither for the healthy controls nor for the pwMS for both the positive and negative contrasts. For the healthy control group the combined 〖OM〗_(v,th) maps at the first scan yielded higher consistency than pwMS group (0.74 and 0.87 consistency peaks vs 0.59 and 0.70 respectively for positive and negative contrast). Considering the longitudinal evaluation instead opposite trends were observed when comparing the 1st and 2nd scans for healthy controls and pre- and post-rehabilitation scans for pwMS. Specifically, a negative trend (decrease) was observed for the healthy control comparing consistency peaks for both positive and negative stimuli. An increase was instead observed when comparing the bar-charts of consistency peaks pre- and post-rehabilitation for pwMS for both positive and negative stimuli. Assessment of inter-subject variability of an ecological social cognition task – Chapter 5 The ESCoT task was effective in eliciting ToM neural networks exclusively when considering explicit ToM reasoning (i.e., answering to questions). A differential neural pattern for affective and cognitive ToM reasoning were distinguished specifically pertaining the Core ToM network (i.e., superior temporal sulcus and temporoparietal junction). When compared to standard GLM derived maps the overlap maps showed a meaningful activation pattern also considering the implicit ToM reasoning. Discussion This PhD project addressed some of the major challenges related to the use of fMRI in longitudinal and rehabilitative frameworks. Specifically, the restrictiveness of fMRI acquisition environment, reducing patients’ compliance to periodical measures, and the intra- and inter- subject variability potentially biasing the interpretation of repeated fMRI measures were specifically addressed. The surface-based approach proposed in Chapter 2 for fNIRS fMRI data integration allowed a straightforward (i.e., vertex-wise) and quantitative comparison between the two techniques. Good results were obtained when considering quantitative spatial and temporal comparisons. Future studies will extend this integration to social cognition tasks in order to fully exploit the ecological acquisition setting offered by fNIRS. The creation of two parallel forms of the same emotional task proposed in Chapter 3 allowed to comprehensively evaluate the mechanism of intra-subject variability at repeated measures. Specifically, it allowed to exclude neural suppression as the predominant effect causing low signal reliability, which might imply instead a change in the neural strategies employed by the subjects to cope with emotional stimuli over time. In Chapter 4 the inter-subject variability of the same emotional paradigm was investigated both cross-sectionally and longitudinally in healthy controls and in pwMS which underwent an EMDR treatment. Opposite trends for healthy controls (decreased consistency) and pwMS (increased consistency) were observed considering the derived overlap-maps when comparing peak bar-charts of consistency longitudinally. The increase in consistency of brain activation in patients’ cohort could be due to a partial restoration of coherence and homogeneity in group neural activation. Furthermore, the neural results are paralleled by a statistically significant improvement in clinical depression symptoms after rehabilitation in response to negative stimuli. Finally, the ESCoT fMRI paradigm implemented in Chapter 5, was effective in eliciting brain activity within ToM circuits. Strong and wider neural activation patterns observed pertaining to explicit ToM, and specifically answering to closed-ended questions, which could be related to a more homogeneous activation across individuals, which was instead reduced for open question and movie. The overlap-maps employed to assess the inter-subject were effective to observe meaningful activation patterns also for implicit ToM reasoning and open-ended questions, despite being characterized by a higher level of individual variability. Overall this PhD project addressed some of the methodological challenges hindering the use of fMRI in longitudinal/rehabilitative settings. Both the implementation of careful task, and complemental analysis methods other than standard statistical GLM approach are warranted to provide a comprehensive depiction of brain function and of the mechanisms at the basis of rehabilitation driven recovery.
Background e Obiettivi La completa comprensione dei meccanismi neurali che determinano la plasticità cerebrale è fondamentale per l'implementazione di approcci su misura che promuovano il recupero da condizioni neurologiche e il ripristino delle funzioni cognitive. Pur essendo il gold standard per la mappatura della funzione cerebrale, l'uso di fMRI in contesti longitudinali è attualmente caratterizzato da diverse sfide che vanno dall’ambiente di acquisizione restrittivo alla variabilità che influisce sulle misure ripetute. In questo contesto, il presente progetto di dottorato, sviluppato presso il Laboratorio CADiTeR dell’IRCCS Fondazione Don Carlo Gnocchi a Milano, affronta alcune di queste sfide per aprire la strada all'impiego della fMRI nell'ambito della riabilitazione clinica, e condurre all'identificazione di potenziali biomarcatori del ripristino della funzione cerebrale. In particolare, questo progetto si prefigge due obiettivi principali: 1) l'integrazione della fMRI con la fNIRS, la quale offre un ambiente di acquisizione meno restrittivo e più ecologico, particolarmente adatto per la valutazione longitudinale e le misurazioni multiple; 2) l'implementazione di paradigmi acquisizione accurati e nuovi metodi per l’analisi dei dati di fMRI, che tengano conto della valutazione della variabilità sia intra- che inter-soggetto che caratterizza le misure di fMRI. Metodi Integrazione di fMRI e fNIRS e valutazione della riproducibilità – Capitolo 2 Diciotto volontari sani sono stati sottoposti ad acquisizioni separate di fMRI e fNIRS, durante le quali hanno eseguito lo stesso paradigma motorio, alternando movimenti di presa di mano destra e sinistra. I dati fMRI sono stati preprocessati utilizzando FreeSurfer FS-FAST e successivamente importati in Brainstorm, utilizzando script sviluppati dall'utente per eseguire l'integrazione con i dati fNIRS. Il set di dati fNIRS è stato preprocessato utilizzando una pipeline standardizzata implementata nel pacchetto Nirstorm di Brainstorm. Le ROI funzionali limitate anatomicamente sono state derivate dalle mappe statistiche sogliate di fNIRS e fMRI (punc <0,05, punc <0,001, pFDR <0,05) all'interno di aree motorie bilaterali (giri pre- e post-centrali) segmentate secondo l'atlante Desikan-Killany. La corrispondenza spaziale e temporale tra le due tecniche sono state valutate calcolando il coefficiente di Dice (DC) e la correlazione di Pearson rispettivamente. Affidabilità e riproducibilità di un paradigma fMRI di stimolazione delle emozioni – Capitolo 3 Sessantadue soggetti sani sono stati sottoposti a due acquisizioni fMRI durante l'acquisizione di un task di stimolazione di emozioni implementato in due forme parallele composte da stimoli visivi diversi ma equivalenti. I dati di fMRI sono stati preprocessati secondo una pipeline standard in SPM-12 e sono stati derivati due diversi contrasti. L'affidabilità del task al test-retest e la riproducibilità (o l'equivalenza) delle forme parallele sono state testate secondo analisi voxel-wise, utilizzando il modello lineare generale (GLM), e attraverso un approccio basato sulla regione di interesse (ROI), calcolando il coefficiente di correlazione intra-classe (ICC) sui contrasti ottenuti. Valutazione della variabilità inter-soggetto di un task di stimolazione delle emozioni: applicazione a coorte di pazienti SM – Capitolo 4 Per questo studio sono stati reclutati trenta volontari sani e tredici pazienti con sclerosi multipla (pwMS) con sintomi depressivi come comorbidità associata a SM. I pazienti MS sono stati sottoposti ad un intervento di riabilitazione EMDR (eye-movement desensitization and reprocessing). Il gruppo di controlli sani è stato sottoposto a due acquisizioni fMRI separate, in cui sono state somministrate due forme parallele del medesimo task di stimolazione di emozione, rispecchiando il setting sperimentale adottate per i pazienti con SM, acquisiti analogamente due volte (prima e dopo la riabilitazione) con due diverse forme parallele. I dati fMRI sono stati preprocessati utilizzando una pipeline standard in SPM-12 e la statistica di gruppo standard GLM. Inoltre, le mappe-T normalizzate sono state derivate per ogni contrasto a livello individuale ed utilizzate per ricavare delle mappe di sovrapposizione ponderata per soglia statistica (threshold-weighted overlap maps) che valutano la variabilità di attivazione tra soggetti. Sia le mappe di attivazione ricavate da GLM che le overlap-maps sono state mascherate utilizzando una maschera inclusiva comprendente 16 ROI bilaterali salienti per il compito emotivo. Valutazione della variabilità inter-soggetto di un task ecologico di cognizione sociale – Capitolo 5 Trentacinque controlli sani sono stati sottoposti a un'acquisizione fMRI durante la somministrazione del test ESCoT adattato a paradigma fMRI. Il task è stato implementato considerando due condizioni: la condizione sperimentale, che valuta la teoria della mente (ToM), versus la condizione di controllo, che valuta l’inferenza fisica (PI); entrambe le condizioni includevano ragionamento implicito (durante la visione dei filmati) ed esplicito (durante la risposta alle domande). SPM-12 è stato utilizzato per il preprocessing dei dati fMRI, seguendo una pipeline standard. A livello dei singoli soggetti sono stati derivati 7 diversi contrasti di confronto tra il ragionamento ToM (implicito/esplicito) e PI. A livello di gruppo le analisi statistiche sono state effettuate tramite GLM e complementate con la creazione di mappe di overlap per la valutazione della variabilità tra soggetti. Entrambe queste analisi sono state eseguite considerando un approccio teorico-guidato utilizzando una maschera inclusiva derivata dall'atlante che rappresenta il modello di Abu-Akel & Shamay-Tsoory. Risultati Integrazione di fMRI e fNIRS e valutazione della riproducibilità – Capitolo 2 Il nuovo approccio di integrazione implementato ha permesso un confronto diretto tra fMRI e fNIRS, sia in termini di valutazione della corrispondenza spaziale che temporale. A livello di gruppo, la corrispondenza spaziale, misurata con il DC, variabile da sostanziale a lieve passando da soglie statistiche meno restrittive punc< 0.05 a più restrittive punc< 0.001. Risultati analoghi sono stati osservati anche a livello individuale. Una correlazione variabile tra moderata e buona è stata osservata globalmente a livello di gruppo per quanto riguarda il confronto temporale. Affidabilità e riproducibilità di un paradigma fMRI di stimolazione delle emozioni – Capitolo 3 Non sono state rilevate differenze significative per le analisi di affidabilità (test-retest) e riproducibilità utilizzando l'analisi voxel-wise bastata su GLM né per il contrasto positivo né per quello negativo. Considerando invece l'approccio basato sulle ROI, l'affidabilità test-retest è risultata scarsa (ICC<0.4) per il contrasto positivo e variabile tra scarsa ed eccellente (0.4≤ICC≤0.75) per quello negativo. Risultati simili sono stati osservati confrontando le due forme parallele (analisi di riproducibilità) con una stabilità del segnale fMRI complessivamente più elevata per gli stimoli negativi rispetto a quelli positivi. Valutazione della variabilità inter-soggetto di un task di stimolazione delle emozioni: applicazione a coorte di pazienti SM – Capitolo 4 Non sono state rilevate differenze significative (pFWE < 0.05) al confronto tra la prima e la seconda scansione di fMRI, né per i controlli sani né per il pwMS sia per i contrasti positivi che negativi. Per il gruppo di controlli sani, le overlap-maps relative alla prima acquisizione hanno rivelato consistenza superiore rispetto al gruppo pwMS (picchi di consistenza di 0.74 e 0.87 vs 0.59 e 0.70 rispettivamente per contrasto positivo e negativo). Considerando la valutazione longitudinale invece le sono stati osservati trend opposto nel confronto tra prima e seconda acquisizione per i controlli sani e pre- e post-riabilitazione per pwMS. In particolare, è stata osservata una diminuzione dei picchi di consistenza nei controlli sani sia per gli stimoli positivi che negativi. Un aumento dei picchi di consistenza è stato invece osservato pre e post riabilitazione per i pazienti con SM sia per stimoli positivi che negativi. Valutazione della variabilità inter-soggetto di un task ecologico di cognizione sociale – Capitolo 5 Il task ESCoT è stato efficace nell’attivazione dei circuiti ToM considerando il ragionamento esplicito (la risposta alle domande). Diversi modelli neurali sono stati distinti per il ragionamento ToM affettivo e cognitivo specificamente relativo alla rete ‘Core ToM’ relativamente al solco temporale superiore e giunzione temporo-parietale. Rispetto alle mappe di attivazione derivate da GLM, le mappe di overlap hanno permesso di apprezzare attivazioni significativo anche considerando il ragionamento ToM implicito. Discussione Questo progetto di dottorato ha affrontato alcune delle principali sfide relative all'uso della fMRI in contesti longitudinali e riabilitativi. In particolare, sono stati affrontati specificamente la restrittività dell’ambiente di acquisizione offerto dall’fMRI, che inficia la complianza dei pazienti a misure periodiche, e la variabilità intra- e inter-soggetto che potrebbe falsare l'interpretazione dei dati fMRI a misure ripetute. L'approccio di integrazione tra fMRI e fNIRS proposto nel capitolo 2 ha permesso un confronto diretto (vertex-wise) e quantitativo tra le due tecniche. Entrambi i confronti quantitativi, spaziali e temporali, hanno rivelato livelli di similarità soddisfacenti. Studi futuri estenderanno il medesimo metodo di integrazione a task di cognizione sociale, al fine di sfruttare appieno il setting di acquisizione ecologico offerta dalla fNIRS. La creazione di due forme parallele di uno stesso task emotivo proposto nel capitolo 3 ha permesso di valutare approfonditamente il meccanismo della variabilità intra-soggetto a misure ripetute. In particolare, ha permesso di escludere la soppressione neurale come effetto predominante che causa bassa affidabilità del segnale, che potrebbe invece essere dovuto ad un cambiamento nelle strategie neurali impiegate dai soggetti per far fronte agli stimoli emotivi nel tempo. Nel capitolo 4 è stata studiata la variabilità inter-soggetto dello stesso paradigma emotivo sia trasversalmente che longitudinalmente nei controlli sani e in pazienti con SM sottoposti a trattamento EMDR. Tendenze opposte per i controlli sani (diminuzione di consistenza) e pwMS (aumento di consistenza) sono stati osservati longitudinalmente considerando i picchi di consistenza derivati dalle mappe di overlap. L'aumento della consistenza dell'attivazione cerebrale nella coorte dei pazienti potrebbe essere dovuto ad un parziale ripristino di coerenza e omogeneità nell'attivazione neurale di gruppo. Inoltre, i risultati neurali sono supportati da un miglioramento statisticamente significativo dei sintomi clinici della depressione dopo la riabilitazione. Infine, il paradigma ESCoT fMRI implementato nel capitolo 5, è stato efficace nell'indurre l'attività cerebrale all'interno dei circuiti ToM. I pattern di attivazione neurale più ampi e robusti osservati relativamente al ragionamento ToM esplicito, potrebbero essere dovuti ad un'attivazione più omogenea tra gli individui, che invece è ridotta nel ragionamento esplicito. Le mappe di overlap impiegate per valutare la variabilità inter-soggetto hanno consentito di osservare modelli di attivazione significativi anche per il ragionamento ToM implicito, nonostante mostrino un livello più elevato di variabilità individuale. Nel complesso questo progetto di dottorato ha affrontato alcune delle sfide metodologiche che ostacolano l'uso della fMRI in ambienti longitudinali/riabilitativi. Implementazioni più accorte dei paradigmi di fMRI, e metodi di analisi complementari, diversi dal l'approccio statistico GLM standard, sono necessari per ottenere una comprensione completa della funzione cerebrale e dei meccanismi alla base del recupero a seguito della riabilitazione.
Towards the use of fMRI in longitudinal protocols to assess rehabilitation-driven neuroplasticity
Pirastru, Alice
2024/2025
Abstract
Background and Aim A comprehensive understanding of the neural mechanisms driving brain plasticity is pivotal for implementing tailored approaches promoting recovery from neurological conditions and restoring cognitive functions. Despite being the gold standard for mapping brain function, the use of fMRI in the longitudinal framework is currently characterized by several challenges ranging from the restrictive acquisition setting to the variability affecting the measures. In this context, this PhD project, developed at the CADiTeR Laboratory of IRCCS Fondazione Don Carlo Gnocchi in Milan, addresses some of these challenges to pave the way for the use of fMRI in the clinical rehabilitative setting, potentially leading to the identification of robust imaging derived biomarkers of restored brain function. Specifically, this project encompasses two major aims: 1) the integration of fMRI with fNIRS modality, offering a less restrictive and more ecological acquisition setting particularly suitable for longitudinal evaluation and multiple measurements; 2) the implementation of careful fMRI paradigm and analysis methods to take into account the evaluation of both intra- and inter-subject variability characterizing fMRI measurements. Methods fMRI-fNIRS Integration and Reliability Assessment – Chapter 2 Eighteen healthy volunteers underwent separate fMRI and fNIRS acquisition performing the same motor (hand grasping) paradigm, alternating right and left hand movements. The functional MRI data were preprocessed using FreeSurfer FS-FAST and were successively imported in Brainstorm, using user-defined scripts to perform integration with fNIRS data. The fNIRS dataset was preprocessed using standardized pipeline implemented in Nirstorm package of Brainstorm. Anatomicallly constrained functional ROIs were derived by thresholding fNIRS and fMRI statistical maps (punc <0.05, punc <0.001, pFDR <0.05) within bilateral motor parcels (pre and post central gyri) segmented according to the Desikan-Killany atlas. Spatial and temporal correspondence between the two techniques was assessed computing Dice Coefficient (DC) and Pearson’s Correlation respectively. Reliability and Reproducibility of an Emotion Generation Task fMRI Paradigm – Chapter 3 Sixty-two healthy subjects were enrolled in the study and performed two fMRI acquisition during the acquisition of an emotion stimulation fMRI task paradigm implemented in two forms including different but equivalent visual stimuli. fMRI data were preprocessed using a standard pipeline in SPM-12. Two-different contrasts were derived. Test-retest reliability and reproducibility (i.e. the equivalence of the parallel forms) of fMRI responses and patterns were evaluated. The analyses were performed voxel-wise, using the general linear model (GLM), and via a region-of-interest (ROI)-based approach, by computing the intra-class correlation coefficient (ICC) on the obtained contrasts. Assessment of inter-subject variability of an emotion stimulation task: application on a cohort of MS patients – Chapter 4 Thirty healthy volunteers and thirteen patients with multiple sclerosis (pwMS), showing depression symptoms as MS-associated comorbidity, were enrolled. pwMS were treated according to an eye-movement desensitization and reprocessing (EMDR) rehabilitation intervention. The healthy control group underwent MRI examination during two separate acquisitions, in which two different parallel forms of the same emotion stimulation task were delivered, mirroring the experimental settings adopted for the pwMS, which were scanned twice before and after rehabilitation. fMRI data were preprocessed using a standard pipeline in SPM-12 and standard GLM group statistic. Furthermore, the normalized T-maps were derived for each contrast at the individual level and used to derive threshold-weighted overlap map (〖OM〗_(thr-w)) assessing inter-subject variability. Both GLM activation maps and OM were masked using a user-defined inclusive mask comprising 16 bilateral ROIs salient for the emotional task. Assessment of inter-subject variability of an ecological social cognition task – Chapter 5 Thirty-five healthy controls underwent an fMRI acquisition during the administration of an adaptation of the ESCoT behavioural test. The task was implemented considering two conditions: the experimental Theory of Mind (ToM) condition versus the control physical inference (PI) condition, each including implicit mentalizing (movie watching), and explicit mentalizing (questions answering). SPM 12 was used for fMRI preprocessing following a standard pipeline. Seven contrasts comparing ToM (implicit/explicit) vs PI reasoning were derived. At group level standard GLM statistics were derived and further complemented by deriving overlap-maps to evaluate inter-subject variability. Both these analyses were performed considering a theory-driven approach using an atlas-derived inclusive mask depicting the Abu-Akel & Shamay-Tsoory model. Results fMRI-fNIRS Integration and Reliability Assessment – Chapter 2 The new implemented integration approach allowed a direct vertex-wise comparison between fMRI and fNIRS both considering a spatial and temporal correspondence assessment. At the group level the spatial correspondence, measured using DC, spanned from substantial to slight agreement moving from punc< 0.05 as the least restrictive to punc< 0.001 as the most conservative threshold. Similar results were also observed at individual level. A moderate to good correlation was observed overall at the group level concerning the temporal comparison. Reliability and Reproducibility of an Emotion Generation Task fMRI Paradigm – Chapter 3 No significant differences for both reliability and reproducibility analyses were revealed using voxel-wise GLM analysis neither for positive nor negative contrast. Considering instead the ROI-based approach, the test-retest reliability was poor (ICC<0.4) for the positive contrast and ranged from poor to excellent (0.4≤ICC≤0.75) for the negative one. Similar results were observed comparing the two different parallel forms (reproducibility analysis) with overall higher fMRI signal stability for negative stimuli over positive ones. Assessment of inter-subject variability of an emotion stimulation task: application on a cohort of MS patients – Chapter 4 No significant differences (pFWE< 0.05) were detected when comparing the 1st and 2nd scans, neither for the healthy controls nor for the pwMS for both the positive and negative contrasts. For the healthy control group the combined 〖OM〗_(v,th) maps at the first scan yielded higher consistency than pwMS group (0.74 and 0.87 consistency peaks vs 0.59 and 0.70 respectively for positive and negative contrast). Considering the longitudinal evaluation instead opposite trends were observed when comparing the 1st and 2nd scans for healthy controls and pre- and post-rehabilitation scans for pwMS. Specifically, a negative trend (decrease) was observed for the healthy control comparing consistency peaks for both positive and negative stimuli. An increase was instead observed when comparing the bar-charts of consistency peaks pre- and post-rehabilitation for pwMS for both positive and negative stimuli. Assessment of inter-subject variability of an ecological social cognition task – Chapter 5 The ESCoT task was effective in eliciting ToM neural networks exclusively when considering explicit ToM reasoning (i.e., answering to questions). A differential neural pattern for affective and cognitive ToM reasoning were distinguished specifically pertaining the Core ToM network (i.e., superior temporal sulcus and temporoparietal junction). When compared to standard GLM derived maps the overlap maps showed a meaningful activation pattern also considering the implicit ToM reasoning. Discussion This PhD project addressed some of the major challenges related to the use of fMRI in longitudinal and rehabilitative frameworks. Specifically, the restrictiveness of fMRI acquisition environment, reducing patients’ compliance to periodical measures, and the intra- and inter- subject variability potentially biasing the interpretation of repeated fMRI measures were specifically addressed. The surface-based approach proposed in Chapter 2 for fNIRS fMRI data integration allowed a straightforward (i.e., vertex-wise) and quantitative comparison between the two techniques. Good results were obtained when considering quantitative spatial and temporal comparisons. Future studies will extend this integration to social cognition tasks in order to fully exploit the ecological acquisition setting offered by fNIRS. The creation of two parallel forms of the same emotional task proposed in Chapter 3 allowed to comprehensively evaluate the mechanism of intra-subject variability at repeated measures. Specifically, it allowed to exclude neural suppression as the predominant effect causing low signal reliability, which might imply instead a change in the neural strategies employed by the subjects to cope with emotional stimuli over time. In Chapter 4 the inter-subject variability of the same emotional paradigm was investigated both cross-sectionally and longitudinally in healthy controls and in pwMS which underwent an EMDR treatment. Opposite trends for healthy controls (decreased consistency) and pwMS (increased consistency) were observed considering the derived overlap-maps when comparing peak bar-charts of consistency longitudinally. The increase in consistency of brain activation in patients’ cohort could be due to a partial restoration of coherence and homogeneity in group neural activation. Furthermore, the neural results are paralleled by a statistically significant improvement in clinical depression symptoms after rehabilitation in response to negative stimuli. Finally, the ESCoT fMRI paradigm implemented in Chapter 5, was effective in eliciting brain activity within ToM circuits. Strong and wider neural activation patterns observed pertaining to explicit ToM, and specifically answering to closed-ended questions, which could be related to a more homogeneous activation across individuals, which was instead reduced for open question and movie. The overlap-maps employed to assess the inter-subject were effective to observe meaningful activation patterns also for implicit ToM reasoning and open-ended questions, despite being characterized by a higher level of individual variability. Overall this PhD project addressed some of the methodological challenges hindering the use of fMRI in longitudinal/rehabilitative settings. Both the implementation of careful task, and complemental analysis methods other than standard statistical GLM approach are warranted to provide a comprehensive depiction of brain function and of the mechanisms at the basis of rehabilitation driven recovery.File | Dimensione | Formato | |
---|---|---|---|
PhD_Thesis_Pirastru_Alice.pdf
solo utenti autorizzati a partire dal 06/12/2025
Descrizione: PhD Thesis Pirastru Alice
Dimensione
7.34 MB
Formato
Adobe PDF
|
7.34 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/10589/232092