Amyotrophic Lateral Sclerosis (ALS) is a neurodegenerative disease highly debilitating with a mean prognosis of a couple of years. Causes are still unknown, and a cure is yet to be discovered. The thesis is centered on the develoopement of some models based on dynamic Baeysian networks, which simulate the progression of the disease from a set of relevant clinical variables collected from the patient. The models, used as a predictive tool to obtain simulations of the progression of the disease could offer clinicians an assistive tool to predict the timeline of the disease for a specific patient, specifying both the times of an impairment and its functional domain, based on MITOS or ALSFRS-R clinical scores. At the same time these models could be used as a framework for in silico simulations to reduce the number of patients in the control group, to which a placebo or no treatment is offered, potentially depriving them of a cure. This thesis is structured in two sections: in the first one, after a brief introduction of the theo- ry relevant to dynamic Bayesian networks, a model from the literature is presented in which the theory is applied to model ALS, and it’s use is studied to produce simulations using visits follo- wing the first one, to evaluate wether using the latest data from the patient could be advantageous for the clinician. In the second part a new network based on ALSFRS-R scale is proposed, implemented in order to use the functional rating scale actually used in clinical practice.
La Sclerosi Laterale Amiotrofica (SLA) è una malattia neurodegenerativa altamente debilitante dalla prognosi media di qualche anno. Le cause non sono ancora del tutto comprese, e si è ancora in cerca di una cura. In questa tesi si sono sviluppati dei modelli basati su reti Bayesiane dinamiche che simulano l’andamento della malattia a partire da una serie di variabili cliniche rilevanti dei pazienti. I modelli, utilizzati a scopo predittico per ottenere delle simulazioni della progressione della malattia, si propongono di offrire in ambito clinico una possibile timeline dell’evoluzione del paziente, prevedendo i momenti delle principali ricadute così come il loro dominio funzionale, come descritto dalle scale cliniche MITOS o ALSFRS-R . Allo stesso tempo questi modelli potrebbero offrire un framework utile per future simulazioni in silico, con le quali confrontare l’andamento di cure sperimentali riducendo il numero di pazienti inseriti nel gruppo di controllo, potenzialmente privati di un medicamento efficace. Questa tesi è strutturata in due parti: nella prima, dopo aver introdotto la teoria relativa alle reti Bayesiane dinamiche, si presenta un modello di letteratura in cu tale metodologia viene uti- lizzata per modellizzare la ALS e se ne studia l’uso per produrre simulazioni, utilizzando come input visite successive alla prima, al fine di valutare l’utilità di utilizzare gli ultimi dati clinici disponibili per stimare l’andamento della malattia. Nella seconda parte si presenta una rete alternativa sviluppata in questa tesi, basata su raggruppa- menti degli score ALSFRS-R, implementata al fine di ottenere una descrizione dell’andamento della malattia che faccia uso della scala di valutazione attualmente adottata nella pratica clinica.
Modelli Bayesiani Dinamici della progressione nella Sclerosi Laterale Amiotrofica
SILVAGNI, LEONARDO
2023/2024
Abstract
Amyotrophic Lateral Sclerosis (ALS) is a neurodegenerative disease highly debilitating with a mean prognosis of a couple of years. Causes are still unknown, and a cure is yet to be discovered. The thesis is centered on the develoopement of some models based on dynamic Baeysian networks, which simulate the progression of the disease from a set of relevant clinical variables collected from the patient. The models, used as a predictive tool to obtain simulations of the progression of the disease could offer clinicians an assistive tool to predict the timeline of the disease for a specific patient, specifying both the times of an impairment and its functional domain, based on MITOS or ALSFRS-R clinical scores. At the same time these models could be used as a framework for in silico simulations to reduce the number of patients in the control group, to which a placebo or no treatment is offered, potentially depriving them of a cure. This thesis is structured in two sections: in the first one, after a brief introduction of the theo- ry relevant to dynamic Bayesian networks, a model from the literature is presented in which the theory is applied to model ALS, and it’s use is studied to produce simulations using visits follo- wing the first one, to evaluate wether using the latest data from the patient could be advantageous for the clinician. In the second part a new network based on ALSFRS-R scale is proposed, implemented in order to use the functional rating scale actually used in clinical practice.File | Dimensione | Formato | |
---|---|---|---|
Silvagni_Leonardo.pdf
embargo fino al 19/07/2027
Dimensione
2.45 MB
Formato
Adobe PDF
|
2.45 MB | Adobe PDF |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/68828