Tasks undertaken during this research deal with development, implementation and testing of a toolbox named TREAT (Tumor gRowth rEsponse to rAdioTherapy), which is applicable for (i) the simulation of solid tumor growth, and (ii) decision making and suggesting different possible treatment plans. To this end, we rely on a mathematical macromodel [Belfatto et al., 2017] of estimating tumor volume growth that takes into account the interplay among tumor volume and the oxygenation. Implemented mathematical approach is based on a macroscopic model exploiting discrete approximation of differential equations of evaluating tumor growth versus time. Starting from two-dynamics (active and necrotic cells) systems, other mechanisms such as the effect of the oxygenation or the occurrence of edema, are also included in the model for some specific application. The proposed simulation model is implemented as a toolbox in MATLAB environment within Graphical User Interface (GUI). TREAT includes two main modules of learning (for optimizing the macromodel based on available clinical measurements) and simulation (for predicting behavior of the patient under a specific treatment plan). For the optimization of the model four different techniques of (i) Gradient-based method (GM), (ii) Monte Carlo (MC), (iii) Genetic Algorithm (GA), and (iv) Imperialist Competitive Algorithm (ICA) are impelemented in TREAT environment. We trained and tested TREAT for an exemplary test case on a prostate tumor-bearing rats dataset including treatment under different irradiation doses. Trained model is then used for the estimation of the tumor growth under a specific (e.g. dose and schedule of radiography) treatment plan. Results show reasonable capabilities of TREAT in reproducing direct clinical measurements. We also evaluated uncertainty of macromodel parameter estimations through Maximum likelihood (ML) framework, and then, evaluated the uncertainties in the evaluation of tumor growth. Results show that using the calibrated simulation model for the estimation of the rat tumor weight after 23 days of treatment (under of 200 R radiotherapy) involves 15.6% uncertainty which is a quite satisfying estimation. Such uncertainty analysis can provide an oncologist extra information about capabilities of the calibrated model in estimating tumor growth behavior while suggesting a patient-individualized treatment plan.
Il lavoro intrapreso durante questa ricerca riguardano lo sviluppo, l'implementazione e il collaudo di una cassetta dei toolbox denominata TREAT (Tumor gRowth rEsponse to rAdionTherapy), applicabile per (i) la simulazione della crescita del tumore solido e (ii) il processo decisionale e il suggerimento di molteplici piani di trattamento. A tal fine, ci basiamo su un macromodel matematico [Belfatto et al., 2017] per stimare la crescita del volume del tumore che tiene conto dell'interazione tra il volume del tumore e l'ossigenazione. L'approccio matematico implementato si basa su un modello macroscopico che sfrutta l'approssimazione discreta di equazioni differenziali per valutare la crescita del tumore in funzione del tempo. A partire da sistemi a due dinamiche (cellule attive e necrotiche), altri modelli come l'effetto dell'ossigenazione o l'insorgenza di edema sono inclusi nel modello per alcune applicazioni specifiche. Il modello di simulazione proposto è implementato come toolbox in ambiente MATLAB all'interno di Graphical User Interface (GUI). TREAT include due moduli principali di learning (per l'ottimizzazione del macromodel basato sulle misurazioni cliniche disponibili) e la simulazione (per predire il comportamento del paziente nell'ambito di un piano di trattamento specifico). Per l'ottimizzazione del modello vengono utilizzate quattro diverse tecniche di (i) Gradient-based method (GM), (ii) Monte Carlo (MC), (iii) Genetic Algorithm (GA), e (iv) Imperialist Competitive Algorithm (ICA) impelementato in ambiente TREAT. Abbiamo addestrato e testato TREAT per un caso esemplare di prova su un set di dati relativi a ratti affetti da un tumore prostatico, incluso il trattamento con diverse dosi di irradiazioni. Il modello addestrato viene quindi utilizzato per la stima della crescita del tumore nel contesto di un piano di trattamento specifico (ad esempio dose e programma di radiografia). I risulti mostrano capacità ragionevoli di TREAT nel riprodurre misure cliniche dirette. Abbiamo anche valutato l'incertezza delle stime dei parametri del macromodello attraverso il framework di Maximum likelihood (ML), e quindi, valutato le incertezze nella valutazione della crescita del tumore. I risultati mostrano che l'uso del modello di simulazione calibrato per la stima del peso del tumore del ratto dopo 23 giorni di trattamento (utilizzando una radioterapia da 200 R) comporta un'incertezza del 15,6% che è una stima abbastanza soddisfacente. Tale analisi di incertezza può fornire ad un oncologo informazioni supplementari sulle capacità del modello calibrato nella stima del comportamento di crescita del tumore, e suggerire un piano di trattamento specifico per il paziente in questione.
Simulation of radiation therapy effects on tumor growth toward a personalized treatment planning : TREAT toolbox
RANAEE, IMAN
2018/2019
Abstract
Tasks undertaken during this research deal with development, implementation and testing of a toolbox named TREAT (Tumor gRowth rEsponse to rAdioTherapy), which is applicable for (i) the simulation of solid tumor growth, and (ii) decision making and suggesting different possible treatment plans. To this end, we rely on a mathematical macromodel [Belfatto et al., 2017] of estimating tumor volume growth that takes into account the interplay among tumor volume and the oxygenation. Implemented mathematical approach is based on a macroscopic model exploiting discrete approximation of differential equations of evaluating tumor growth versus time. Starting from two-dynamics (active and necrotic cells) systems, other mechanisms such as the effect of the oxygenation or the occurrence of edema, are also included in the model for some specific application. The proposed simulation model is implemented as a toolbox in MATLAB environment within Graphical User Interface (GUI). TREAT includes two main modules of learning (for optimizing the macromodel based on available clinical measurements) and simulation (for predicting behavior of the patient under a specific treatment plan). For the optimization of the model four different techniques of (i) Gradient-based method (GM), (ii) Monte Carlo (MC), (iii) Genetic Algorithm (GA), and (iv) Imperialist Competitive Algorithm (ICA) are impelemented in TREAT environment. We trained and tested TREAT for an exemplary test case on a prostate tumor-bearing rats dataset including treatment under different irradiation doses. Trained model is then used for the estimation of the tumor growth under a specific (e.g. dose and schedule of radiography) treatment plan. Results show reasonable capabilities of TREAT in reproducing direct clinical measurements. We also evaluated uncertainty of macromodel parameter estimations through Maximum likelihood (ML) framework, and then, evaluated the uncertainties in the evaluation of tumor growth. Results show that using the calibrated simulation model for the estimation of the rat tumor weight after 23 days of treatment (under of 200 R radiotherapy) involves 15.6% uncertainty which is a quite satisfying estimation. Such uncertainty analysis can provide an oncologist extra information about capabilities of the calibrated model in estimating tumor growth behavior while suggesting a patient-individualized treatment plan.| File | Dimensione | Formato | |
|---|---|---|---|
|
ImanRanaee-Thesis.pdf
solo utenti autorizzati dal 09/07/2022
Descrizione: Thesis
Dimensione
4.46 MB
Formato
Adobe PDF
|
4.46 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/10589/149022