This thesis is part of the GUIDE-AI project, funded by the European Commission. The aim of GUIDE-AI is to leverage artificial intelligence to increase adherence to clinical guidelines in the management of therapeutic pathways for patients with chronic diseases in Europe. Because many patients do not receive the recommended optimal pharmacological therapy, owing to the length and complexity of the guidelines, the project aims to increase adherence to evidence-based therapeutic pathways and thereby reduce the mortality rate attributable to chronic diseases in Europe. GUIDE-AI is being developed as a reference model for four clinical indications: chronic obstructive pulmonary disease, chronic heart failure, chronic kidney disease and asthma. This thesis is situated in the early phases of the GUIDE-AI project and focuses on the analysis of the clinical guideline for chronic obstructive pulmonary disease (COPD), examining the entire therapeutic pathway. The work is organized into four sequential macro-phases. The first phase concerned the formalization of the clinical knowledge contained in the Global Initiative for Chronic Obstructive Lung Disease (GOLD, 2026 report) guideline, translated into a structured flowchart that maps the entire therapeutic pathway, from initial patient intake to follow-up. The second phase involved the formal representation of this knowledge within an expert system, using the PROforma language by means of the Composer software, ensuring deterministic, verifiable behavior that is free of hallucinations. The third phase consists of building the knowledge base on which the project's Retrieval-Augmented Generation (RAG) architecture will subsequently rely, by implementing a Knowledge Graph from the reference guideline. The fourth phase consists of using a pipeline for the automatic extraction of clinical concepts and their mapping to the standardized SNOMED CT nomenclature. This last step is essential to complete the implementation of the Knowledge Graph, as it makes it possible to link the chapters of the guideline to standardized SNOMED CT concepts. Overall, the work demonstrates the importance of using knowledge-based artificial intelligence systems in clinical settings, as well as their feasibility through expert systems and RAG.
Il presente lavoro di tesi si colloca all'interno del progetto GUIDE-AI, finanziato dalla Commissione Europea. L'obiettivo di GUIDE-AI è di sfruttare l'intelligenza artificiale per aumentare l'aderenza alle linee guida cliniche nella gestione dei percorsi terapeutici di pazienti affetti da malattie croniche in Europa. Un numero elevato di pazienti non riceve la terapia farmacologica ottimale raccomandata, anche a causa della lunghezza e della complessità delle linee guida. Il progetto si pone l’obiettivo di incrementare l’aderenza a percorsi terapeutici fondati su evidenze scientifiche e, di conseguenza, di ridurre il tasso di mortalità attribuibile a malattie croniche in Europa. GUIDE-AI viene sviluppato come modello di riferimento per quattro indicazioni cliniche: broncopneumopatia cronica ostruttiva, scompenso cardiaco cronico, malattia renale cronica e asma. Il presente lavoro di tesi si inserisce all'interno delle fasi iniziali del progetto GUIDE-AI e si focalizza sull'analisi della linea guida clinica per la broncopneumopatia cronica ostruttiva (BPCO o Chronic obstructive pulmonary disease, COPD), approfondendone l'intero percorso terapeutico. L'attività si articola in quattro macro-fasi sequenziali. La prima fase è la formalizzazione della conoscenza clinica contenuta nella linea guida Global Initiative for Chronic Obstructive Lung Disease (GOLD, report 2026), tradotta in un flowchart strutturato che mappa l'intero percorso terapeutico, dalla presa in carico iniziale al follow-up. La seconda fase ha previsto la rappresentazione formale di tale conoscenza all'interno di un sistema esperto utilizzando il linguaggio PROforma per mezzo del software Composer, garantendo un comportamento deterministico, verificabile e privo di allucinazioni. La terza fase prevede la costruzione della base di conoscenza su cui poi si baserà l’architettura di Retrieval Augmented Generation (RAG) del progetto, implementando un grafo di conoscenza (Knowledge Graph) con le linee guida di riferimento. La quarta fase prevede l’utilizzo di una pipeline per l'estrazione automatica di concetti clinici e la loro mappatura alla nomenclatura standardizzata SNOMED CT. Quest’ultimo passaggio è essenziale per completare l’implementazione del grafo di conoscenza, rendendo possibile il legame tra i capitoli della linea guida e i concetti standardizzati SNOMED CT. Nel complesso, il lavoro dimostra l’importanza di utilizzare sistemi di intelligenza artificiale basati sulla conoscenza in contesti clinici e la loro fattibilità tramite sistemi esperti e RAG.
Modelli di rappresentazione della conoscenza per il supporto alla decisione clinica in sistemi basati sull’intelligenza artificiale
BUIATTI, SOFIA
2025/2026
Abstract
This thesis is part of the GUIDE-AI project, funded by the European Commission. The aim of GUIDE-AI is to leverage artificial intelligence to increase adherence to clinical guidelines in the management of therapeutic pathways for patients with chronic diseases in Europe. Because many patients do not receive the recommended optimal pharmacological therapy, owing to the length and complexity of the guidelines, the project aims to increase adherence to evidence-based therapeutic pathways and thereby reduce the mortality rate attributable to chronic diseases in Europe. GUIDE-AI is being developed as a reference model for four clinical indications: chronic obstructive pulmonary disease, chronic heart failure, chronic kidney disease and asthma. This thesis is situated in the early phases of the GUIDE-AI project and focuses on the analysis of the clinical guideline for chronic obstructive pulmonary disease (COPD), examining the entire therapeutic pathway. The work is organized into four sequential macro-phases. The first phase concerned the formalization of the clinical knowledge contained in the Global Initiative for Chronic Obstructive Lung Disease (GOLD, 2026 report) guideline, translated into a structured flowchart that maps the entire therapeutic pathway, from initial patient intake to follow-up. The second phase involved the formal representation of this knowledge within an expert system, using the PROforma language by means of the Composer software, ensuring deterministic, verifiable behavior that is free of hallucinations. The third phase consists of building the knowledge base on which the project's Retrieval-Augmented Generation (RAG) architecture will subsequently rely, by implementing a Knowledge Graph from the reference guideline. The fourth phase consists of using a pipeline for the automatic extraction of clinical concepts and their mapping to the standardized SNOMED CT nomenclature. This last step is essential to complete the implementation of the Knowledge Graph, as it makes it possible to link the chapters of the guideline to standardized SNOMED CT concepts. Overall, the work demonstrates the importance of using knowledge-based artificial intelligence systems in clinical settings, as well as their feasibility through expert systems and RAG.| File | Dimensione | Formato | |
|---|---|---|---|
|
Buiatti_Sofia.pdf
accesso aperto
Dimensione
3.73 MB
Formato
Adobe PDF
|
3.73 MB | Adobe PDF | Visualizza/Apri |
È consentito all'utente scaricare e condividere i documenti disponibili a testo pieno in UNITESI UNIPV nel rispetto della licenza Creative Commons del tipo CC BY NC ND.
Per maggiori informazioni e per verifiche sull'eventuale disponibilità del file scrivere a: [email protected].
https://hdl.handle.net/20.500.14239/37283