Quantitative imaging (Q-imaging) represents a powerful tool for non-invasive tumor characterization, providing quantitative imaging biomarkers (QIBs) that extend beyond conventional anatomical assessment. When integrated with computational modeling approaches, these methods enable the characterization of tissue properties across multiple spatial scales, from the macroscopic (voxel-level) to the microscopic (sub-voxel) level, allowing insights into tumor heterogeneity, aggressiveness, and response to therapy. Diffusion-Weighted Magnetic Resonance Imaging (DWI) represents one of the most relevant Q-imaging modalities, as it enables the non-invasive quantification of water molecule mobility, which reflects underlying variations in tissue composition and microstructural heterogeneity. The integration of Q-imaging with multiscale modeling is particularly relevant in the context of radiotherapy (RT) and particle therapy (PT), where spatial and biological heterogeneity significantly impact treatment response. This PhD thesis investigates the integration of Q-imaging and multiscale modeling for biologically informed and personalized approaches in PT. A comprehensive computational framework was developed, combining advanced QIBs, Monte Carlo simulations, and validation strategies encompassing both macroscopic and microscopic scales. At the macroscopic level, radiomics and dosiomics analyses were performed to extract quantitative features from DWI and both biological dose maps and dose-averaged linear energy transfer (LETd) maps, enabling the development of models for tumor characterization and treatment outcome prediction. At the microscopic level, biophysical modeling based on Monte Carlo simulations of water diffusion within synthetic cellular substrates allowed estimation of tissue-specific microstructural parameters from routine DWI data. These models were applied to investigate biological aggressiveness, recurrence risk, and to map intratumoral subregions (habitats) associated with unfavorable outcomes. In addition, a dedicated framework for histological validation was implemented. Combining in vivo MRI and ex vivo imaging with patient-specific 3D-printed molds, this approach enabled accurate alignment between DWI-derived parameters and histological cell density, providing insight into the biological accuracy of different DWI models. Finally, a simulation tool based on the Geant4 Monte Carlo codes was developed to model microscale radiation-matter interactions in realistic multicellular geometries derived from microscopy, providing a framework for future integration between imaging and radiobiological modeling. Overall, this thesis provides a comprehensive framework integrating Q-imaging and multiscale modeling to enhance tumor characterization and guide biologically informed, personalized strategies in particle therapy.
L’imaging quantitativo (Q-imaging) rappresenta una metodologia avanzata per la caratterizzazione non invasiva dei tumori attraverso l’estrazione di biomarcatori (QIBs, Quantitative Imaging Biomarkers) che estendono le capacità di caratterizzazione oltre la valutazione anatomica convenzionale. Integrato con approcci di modellizzazione computazionale, il Q-imaging consente di descrivere le proprietà tissutali su scale spaziali multiple, dal livello macroscopico (voxel) a quello microscopico (sub-voxel), offrendo nuove prospettive sull’eterogeneità tumorale, l’aggressività biologica e sulla risposta alle terapie. La Risonanza Magnetica pesata in Diffusione (DWI) rappresenta una delle tecniche di Q-imaging più rilevanti poiché permette di quantificare in modo non invasivo la mobilità delle molecole d’acqua, che riflette variazioni nella composizione tissutale e nell’eterogeneità microstrutturale. L’integrazione tra Q-imaging e modellizzazione multiscala è particolarmente rilevante nel contesto della radioterapia convenzionale (RT) e della terapia con particelle (PT), dove l’eterogeneità spaziale e biologica influisce in modo significativo sulla risposta al trattamento. Questa tesi di dottorato esplora l’integrazione del Q-imaging con la modellizzazione multiscala per sviluppare approcci personalizzati e biologicamente informati in PT. A tal fine è stato elaborato un framework computazionale completo, che combina QIBs avanzati, simulazioni Monte Carlo e strategie di validazione su scale macroscopiche e microscopiche. Sul piano macroscopico, sono state sviluppate analisi radiomiche e dosiomiche per estrarre caratteristiche quantitative da DWI, da mappe di dose biologica e da mappe di trasferimento lineare di energia (LETd), con l’obiettivo di sviluppare modelli per la caratterizzazione tumorale e la predizione dell’efficacia del trattamento. A livello microscopico, la modellizzazione biofisica basata su simulazioni Monte Carlo della diffusione delle molecole d’acqua in substrati cellulari sintetici ha consentito di stimare parametri microstrutturali specifici dei tessuti a partire da dati DWI di routine. Tali modelli sono stati applicati per studiare l’aggressività tumorale, il rischio di recidiva e per mappare sottoregioni intra-tumorali (habitats) associate a prognosi sfavorevole. Inoltre, è stato sviluppato un framework dedicato alla validazione istologica che, integrando imaging in vivo ed ex vivo con stampi 3D personalizzati per il paziente, ha permesso un allineamento accurato tra i parametri derivati dalla DWI e la densità cellulare istologica, fornendo informazioni sull’accuratezza biologica dei diversi modelli DWI. Infine, è stato sviluppato uno strumento di simulazione basato su codici Monte Carlo (Geant4) per modellizzare le interazioni radiazione-materia su scala microscopica in geometrie multicellulari realistiche derivate da microscopia, fornendo un framework per una futura integrazione tra imaging e modellizzazione radiobiologica. Nel complesso, questa tesi propone un framework integrato di Q-imaging e modellizzazione multiscala finalizzato a migliorare la caratterizzazione dei tumori e a guidare strategie terapeutiche personalizzate e biologicamente guidate nella radioterapia con particelle.
Quantitative imaging and multi-scale modelling in personalized radiotherapy
Morelli, Letizia
2024/2025
Abstract
Quantitative imaging (Q-imaging) represents a powerful tool for non-invasive tumor characterization, providing quantitative imaging biomarkers (QIBs) that extend beyond conventional anatomical assessment. When integrated with computational modeling approaches, these methods enable the characterization of tissue properties across multiple spatial scales, from the macroscopic (voxel-level) to the microscopic (sub-voxel) level, allowing insights into tumor heterogeneity, aggressiveness, and response to therapy. Diffusion-Weighted Magnetic Resonance Imaging (DWI) represents one of the most relevant Q-imaging modalities, as it enables the non-invasive quantification of water molecule mobility, which reflects underlying variations in tissue composition and microstructural heterogeneity. The integration of Q-imaging with multiscale modeling is particularly relevant in the context of radiotherapy (RT) and particle therapy (PT), where spatial and biological heterogeneity significantly impact treatment response. This PhD thesis investigates the integration of Q-imaging and multiscale modeling for biologically informed and personalized approaches in PT. A comprehensive computational framework was developed, combining advanced QIBs, Monte Carlo simulations, and validation strategies encompassing both macroscopic and microscopic scales. At the macroscopic level, radiomics and dosiomics analyses were performed to extract quantitative features from DWI and both biological dose maps and dose-averaged linear energy transfer (LETd) maps, enabling the development of models for tumor characterization and treatment outcome prediction. At the microscopic level, biophysical modeling based on Monte Carlo simulations of water diffusion within synthetic cellular substrates allowed estimation of tissue-specific microstructural parameters from routine DWI data. These models were applied to investigate biological aggressiveness, recurrence risk, and to map intratumoral subregions (habitats) associated with unfavorable outcomes. In addition, a dedicated framework for histological validation was implemented. Combining in vivo MRI and ex vivo imaging with patient-specific 3D-printed molds, this approach enabled accurate alignment between DWI-derived parameters and histological cell density, providing insight into the biological accuracy of different DWI models. Finally, a simulation tool based on the Geant4 Monte Carlo codes was developed to model microscale radiation-matter interactions in realistic multicellular geometries derived from microscopy, providing a framework for future integration between imaging and radiobiological modeling. Overall, this thesis provides a comprehensive framework integrating Q-imaging and multiscale modeling to enhance tumor characterization and guide biologically informed, personalized strategies in particle therapy.| File | Dimensione | Formato | |
|---|---|---|---|
|
PhD_thesis_LetiziaMorelli.pdf
non accessibile
Dimensione
40.86 MB
Formato
Adobe PDF
|
40.86 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/10589/243392