This PhD thesis investigates the application of network science to neuroscience and psychopathology, with the overarching aim of advancing theoretical understanding, methodological innovation, and empirical insights across a wide range of clinical domains. It is the result of three years of research dedicated to illustrating the clinical utility of network science, both in the complex study of populations with psychiatric disorders, in psychometric investigations examining tests and their underlying factorial structures, and in analyses aimed at predicting clinical phenotypes and pharmacological treatment trajectories. The thesis is organised into three main sections. The first part describes the conceptual and methodological foundations, introducing the principles of complex systems and network science and considering their relevance to psychology, psychiatry, and cognitive neuroscience. Particular attention is placed on the transition from latent-variable models to network methods, with emphasis on transdiagnostic and dimensional viewpoints. In addition, the issues of network stability, critical transitions, and hysteresis are considered as they relate to resilience, relapse, and complex system dynamics. The second part comprises empirical applications of network analysis across study designs and clinical samples. Partial correlation networks are applied to represent symptom interactions in depression, bipolar disorder, and caregiver burden in children diagnosed with cancer. Mixed Graphical Models show they are capable of dealing with mixed-type data such that differential risk profiles for suicide attempts, sex-differentiated neuropsychological outcomes, and apathy factors in cognitive decline can be recognised. Exploratory Graph Analysis is applied to structural psychometric investigations of apathy and schizotypy, offering data-driven technique in comparison with alternative factor analytic methods. Network Intervention Analysis addresses longitudinal approaches, demonstrating the dynamic interplay of symptoms and interventions in psychiatry and paediatric oncology, and the development of psychotic disorders under antipsychotic treatment. Finally, similarity networks of patients coupled with machine learning approaches are applied at the level of the patient for classification purposes, with applications in Alzheimer's disease, psychosis, and personalised prediction of clinical severity profiles. Data for all studies presented in this thesis were collected across multiple national and international clinical centres, encompassing diverse populations and diagnostic categories. This multicentric framework provided a unique opportunity to evaluate network-based models across heterogeneous clinical contexts, thereby strengthening the external validity and translational relevance of the findings. The third part reflects on methodological issues and integrative perspectives, critically evaluating network methodology's limitations and assets. In summary, this thesis analyses and extends existing literature, and provides novel evidence demonstrating how network analysis can bridge the gap between complex systems theory and applied neuroscience, by offering innovative methods to study multidimensional interactions, refine nosological models, and foster a model of medicine that is more precise, predictive, preventive, and personalised.
Questa tesi di dottorato indaga l’applicazione della scienza delle reti alle neuroscienze e alla psicopatologia, con l’obiettivo generale di promuovere il progresso teorico, l’innovazione metodologica e l’approfondimento empirico in un’ampia gamma di ambiti clinici. È il risultato di tre anni di ricerca dedicati a illustrare l’utilità clinica della scienza delle reti, sia nello studio complesso delle popolazioni con disturbi psichiatrici, sia nelle indagini psicometriche volte a esaminare i test e le loro strutture fattoriali sottostanti, sia nelle analisi finalizzate a predire fenotipi clinici e traiettorie di trattamento farmacologico. La tesi è organizzata in tre sezioni principali. La prima parte descrive le basi concettuali e metodologiche, introducendo i principi dei sistemi complessi e della scienza delle reti, e considerando la loro rilevanza per la psicologia, la psichiatria e le neuroscienze cognitive. Particolare attenzione è rivolta alla transizione dai modelli a variabili latenti ai metodi basati su reti, con enfasi sugli approcci transdiagnostici e dimensionali. Inoltre, vengono affrontati i temi della stabilità delle reti, delle transizioni critiche e dell’isteresi, in relazione alla resilienza, alla ricaduta e alle dinamiche dei sistemi complessi. La seconda parte comprende applicazioni empiriche dell’analisi di rete in diversi disegni di studio e campioni clinici. Le reti di correlazione parziale vengono applicate per rappresentare le interazioni tra sintomi nella depressione, nel disturbo bipolare e nel burden dei caregiver di bambini affetti da cancro. I Mixed Graphical Models dimostrano di essere in grado di gestire dati di tipo misto, consentendo di identificare profili di rischio differenziale per i tentativi di suicidio, esiti neuropsicologici differenziati per sesso e fattori di apatia nel declino cognitivo. L’Exploratory Graph Analysis viene applicata a indagini psicometriche strutturali su apatia e schizotipia, offrendo una tecnica data-driven in confronto con i metodi di analisi fattoriale tradizionali. La Network Intervention Analysis affronta gli approcci longitudinali, mostrando l’interazione dinamica tra sintomi e interventi in psichiatria e oncologia pediatrica, nonché lo sviluppo dei disturbi psicotici sotto trattamento antipsicotico. Infine, le reti di similarità tra pazienti, combinate con approcci di machine learning, vengono applicate a livello individuale per scopi di classificazione, con applicazioni nella malattia di Alzheimer, nella psicosi e nella predizione personalizzata dei profili di gravità clinica. I dati di tutti gli studi presentati in questa tesi sono stati raccolti in numerosi centri clinici nazionali e internazionali, includendo popolazioni e categorie diagnostiche eterogenee. Questo quadro multicentrico ha offerto un’opportunità unica per valutare i modelli basati su reti in contesti clinici diversificati, rafforzando così la validità esterna e la rilevanza traslazionale dei risultati. La terza parte riflette sulle questioni metodologiche e sulle prospettive integrative, valutando criticamente i limiti e i punti di forza della metodologia delle reti. In sintesi, questa tesi analizza e amplia la letteratura esistente, fornendo nuove evidenze che dimostrano come l’analisi di rete possa colmare il divario tra la teoria dei sistemi complessi e le neuroscienze applicate, offrendo metodi innovativi per studiare le interazioni multidimensionali, affinare i modelli nosologici e promuovere un modello di medicina più preciso, predittivo, preventivo e personalizzato.
L’approccio di rete nelle neuroscienze e nella psicopatologia: dai piccoli mondi ai sistemi complessi
SARTI, PIERFRANCESCO
2026
Abstract
This PhD thesis investigates the application of network science to neuroscience and psychopathology, with the overarching aim of advancing theoretical understanding, methodological innovation, and empirical insights across a wide range of clinical domains. It is the result of three years of research dedicated to illustrating the clinical utility of network science, both in the complex study of populations with psychiatric disorders, in psychometric investigations examining tests and their underlying factorial structures, and in analyses aimed at predicting clinical phenotypes and pharmacological treatment trajectories. The thesis is organised into three main sections. The first part describes the conceptual and methodological foundations, introducing the principles of complex systems and network science and considering their relevance to psychology, psychiatry, and cognitive neuroscience. Particular attention is placed on the transition from latent-variable models to network methods, with emphasis on transdiagnostic and dimensional viewpoints. In addition, the issues of network stability, critical transitions, and hysteresis are considered as they relate to resilience, relapse, and complex system dynamics. The second part comprises empirical applications of network analysis across study designs and clinical samples. Partial correlation networks are applied to represent symptom interactions in depression, bipolar disorder, and caregiver burden in children diagnosed with cancer. Mixed Graphical Models show they are capable of dealing with mixed-type data such that differential risk profiles for suicide attempts, sex-differentiated neuropsychological outcomes, and apathy factors in cognitive decline can be recognised. Exploratory Graph Analysis is applied to structural psychometric investigations of apathy and schizotypy, offering data-driven technique in comparison with alternative factor analytic methods. Network Intervention Analysis addresses longitudinal approaches, demonstrating the dynamic interplay of symptoms and interventions in psychiatry and paediatric oncology, and the development of psychotic disorders under antipsychotic treatment. Finally, similarity networks of patients coupled with machine learning approaches are applied at the level of the patient for classification purposes, with applications in Alzheimer's disease, psychosis, and personalised prediction of clinical severity profiles. Data for all studies presented in this thesis were collected across multiple national and international clinical centres, encompassing diverse populations and diagnostic categories. This multicentric framework provided a unique opportunity to evaluate network-based models across heterogeneous clinical contexts, thereby strengthening the external validity and translational relevance of the findings. The third part reflects on methodological issues and integrative perspectives, critically evaluating network methodology's limitations and assets. In summary, this thesis analyses and extends existing literature, and provides novel evidence demonstrating how network analysis can bridge the gap between complex systems theory and applied neuroscience, by offering innovative methods to study multidimensional interactions, refine nosological models, and foster a model of medicine that is more precise, predictive, preventive, and personalised.| File | Dimensione | Formato | |
|---|---|---|---|
|
PhD thesis_SARTI_a.pdf
embargo fino al 19/02/2027
Licenza:
Tutti i diritti riservati
Dimensione
12.82 MB
Formato
Adobe PDF
|
12.82 MB | Adobe PDF |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14242/358261
URN:NBN:IT:UNIMORE-358261