Hadron therapy is a widely employed technique that uses protons and heavy ions to treat cancer. It has the potential of delivering highly conformal dose distributions to the tumor volume while sparing the surrounding healthy tissue, thanks to the dose distribution characterized by the Bragg peak at the end of charged particles range. In order to exploit the full potential of hadron therapy, an in vivo monitoring technique is desirable in order to reduce the uncertainties and therefore the treatment safety margins. Positron emission tomography (PET) is considered one of the most promising in vivo non-invasive imaging techniques for monitoring the particle range in radiation treatments. One of the data acquisition methods is the so-called in-beam which is performed during irradiation at the treatment site. The problem of in-beam monitoring is that in-spill data are much noisier while inter-spill data for accelerators with high duty cycles, are much less due to the small number of acquired decays. During the spills, the noisy background is due to the presence of strong beam-induced radiation that increases the random coincidence rates. This background might originate from the decay of β+ emitters with half-lives in millisecond range and high endpoint energies, by γ-rays following nuclear reactions not related to β+ decay or by pair productions and neutrons. The noisy events cannot be separated from the usable decays of long-lived β+ emitters and cannot be corrected with standard random coincidence correction techniques because of the time-correlation of the beam-induced background with the ion beam microstructure. Until now, only two methods exist for identifying coincident events that occur during the microbunches in the spills. Both of them use information about the beam microstructure from external sources. In the first method, the RF signal from the accelerator is used externally and the data processing is done offline. In the second one, a fast particle detector placed in the beam path before the target is used and the process is triggered only when a particle arrives. With this thesis, the correlation between the beam microstructure and the RF of the synchrotron is confirmed by analyzing the events in the spills without the need of an external signal. An algorithm for the calculation of the period of the beam microstructure is developed. Small differences in the period between the spills impose the separate analysis for every spill. The period is calculated with 4 digits precision in nanosecond time scale, making a significant difference to the representation of the microbunch. In the end, the firmware related to the algorithm for the calculation of the period of the beam microstructure is developed using only the events in the spills. The simulation results show that it is possible the algorithm to be implemented in an FPGA and provide information about the period of the beam microstructure in real time. Moreover, a coincidence sorter is developed in order to provide real time coincidence detection. The simulation results for the two different architectures of the sorter that uses comparators with two and three inputs, are presented. The 3D spatial distribution and the 1D activity profiles of the coincidence events are constructed for inter-spill and in-spill data. The strong radiation background is visible in the reconstructed images, especially before the entrance surface of the phantom and at the end of the activity range with a tail. After filtering out the in-spill data by discarding the coincidence events that occur in a sub-interval of the microbunch, it is shown that the reconstructed image improves severely. In the 1D activity profile, one can observe that the number of coincidence events before the entrance surface of the phantom decreases significantly. This might happen because neutrons are discarded since they are detected a few ns later after the interaction of the beam with the nuclei. Results show that the signal to noise ratio (SNR), defined as the activity peak in the phantom divided by the background level, is improved by a factor of about 4.8 with respect to the in-spill signal. In the end, it is important to mention that this activity has been developed within the projects INSIDE and INFIERI (FP7-PEOPLE-2012-ITN project number 317446) funded by MIUR and EU respectively.
La terapia adronica è una tecnica che usa protoni e ioni positivi per trattare il cancro. Essa permette di applicare distribuzioni di dose sul tessuto tumorale con precisioni irraggiungibili attraverso altre tecniche, come per esempio quelle radioterapiche convenzionali. Tale precisione fa sì che la tossicità sul tessuto sano circostante venga ridotta al minimo. Al fine di sfruttare a pieno le potenzialità della terapia adronica, è necessario disporre di una tecnica di monitoraggio in-vivo, che permetta di ridurre l'incertezza di range delle particelle e quindi i margini di sicurezza del trattamento. La tomografia ad emissione positronica (PET) è considerata una delle tecniche di imaging non invasivo in-vivo più mature per il monitoraggio del range nei trattamenti radioterapici. Quando l'acquisizione dei dati PET viene eseguita durante l'irraggiamento, il monitoraggio viene indicato come in-beam. Il problema del monitoraggio in-beam è che i dati acquisiti durante l'irraggiamento (in-spill) sono molto rumorosi mentre quelli acquisiti nelle pause (inter-spill) lo sono molto meno. Durante l'inter-spill, tuttavia, il segnale è anche molto debole perché molti degli emettitori β+ non sono stati ancora prodotti. Il rumore di fondo durante l'in-spill è dovuto alla forte radiazione rivelata mentre il fascio colpisce il tessuto bersaglio. Tale radiazione aumenta la probabilità di rivelazione di coincidenze random, a loro volta fonte di rumore nell'imaging PET. La radiazione di fondo può essere causata dal decadimento di emettitori β+ con vita media nell'ordine dei millisecondi, da raggi γ prodotti immediati (prompt) di reazioni nucleari non correlate al decadimento β+, da produzioni di coppie positrone-elettrone o da neutroni. Gli eventi random non possono essere discriminati dai decadimenti β+ utili, e non possono essere corretti con le tecniche convenzionali di stima delle coincidenze random a causa della mancanza di correlazione temporale tra i decadimenti β+ e i prodotti nucleari immediati. Allo stato dell'arte sono stati proposti due metodi per separare le coincidenze in-spill che si verificano durante la fase di estrazione dall'acceleratore da quelle che si verificano durante la fase di accelerazione. Le prime sono naturalmente più rumorose delle seconde, in quanto durante la fase di accelerazione sono assenti gli eventi prompt. Entrambi i metodi utilizzano informazioni sulla microstruttura del fascio da fonti esterne. Nel primo metodo, il segnale RF dall'acceleratore viene utilizzato per generare un segnale di gate all'interno del sistema di acquisizione PET. Nel secondo, il segnale di gate viene prodotto da un rivelatore di particelle inserito nel percorso del fascio prima del bersaglio. In questa tesi, le fasi della microstruttura del fascio vengono rivelate direttamente analizzando la distribuzione temporale degli eventi PET acquisiti in-spill. A tal fine, un nuovo algoritmo è stato sviluppato e validato in simulazione e sperimentalmente per il rilevamento della microstruttura temporale del fascio. I risultati di simulazione mostrano che e possibile implementare l'algoritmo su un FPGA e sfruttare le informazioni da esso prodotte per discriminare le coincidenze avvenute durante le fasi di estrazione da quelle avvenute durante le fasi di accelerazione. La distribuzione spaziale 3D ei profili di attività 1D degli eventi di coincidenza sono ricavati con un algoritmo di ricostruzione tomografica ML-EM per i dati interspill e in-spill. Il rumore di background è visibile nelle immagini ricostruite da dati sperimentali in-spill. Dopo aver filtrato i dati in-spill, scartando gli eventi di coincidenza che si verificano durante la fase di estrazione, si dimostra che l'immagine ricostruita migliora significativamente. Nel profilo di attività 1D si osserva una forte diminuzione della baseline del segnale, corrispondente al contributo delle coincidenze random. In particolare, il rapporto fra il picco di attività nel target diviso per il livello di fondo migliora di un fattore 4.8. Questa attività è stata sviluppata all'interno dei progetti INSIDE e INFIERI (FP7-PEOPLE-2012-ITN project number 317446), finanziati rispettivamente da MIUR e EU.
Kostara, E. (2017). Full-beam PET monitoring in hadron therapy and related coincidence logic.
Full-beam PET monitoring in hadron therapy and related coincidence logic
KOSTARA, ELEFTHERIA
2017-01-01
Abstract
Hadron therapy is a widely employed technique that uses protons and heavy ions to treat cancer. It has the potential of delivering highly conformal dose distributions to the tumor volume while sparing the surrounding healthy tissue, thanks to the dose distribution characterized by the Bragg peak at the end of charged particles range. In order to exploit the full potential of hadron therapy, an in vivo monitoring technique is desirable in order to reduce the uncertainties and therefore the treatment safety margins. Positron emission tomography (PET) is considered one of the most promising in vivo non-invasive imaging techniques for monitoring the particle range in radiation treatments. One of the data acquisition methods is the so-called in-beam which is performed during irradiation at the treatment site. The problem of in-beam monitoring is that in-spill data are much noisier while inter-spill data for accelerators with high duty cycles, are much less due to the small number of acquired decays. During the spills, the noisy background is due to the presence of strong beam-induced radiation that increases the random coincidence rates. This background might originate from the decay of β+ emitters with half-lives in millisecond range and high endpoint energies, by γ-rays following nuclear reactions not related to β+ decay or by pair productions and neutrons. The noisy events cannot be separated from the usable decays of long-lived β+ emitters and cannot be corrected with standard random coincidence correction techniques because of the time-correlation of the beam-induced background with the ion beam microstructure. Until now, only two methods exist for identifying coincident events that occur during the microbunches in the spills. Both of them use information about the beam microstructure from external sources. In the first method, the RF signal from the accelerator is used externally and the data processing is done offline. In the second one, a fast particle detector placed in the beam path before the target is used and the process is triggered only when a particle arrives. With this thesis, the correlation between the beam microstructure and the RF of the synchrotron is confirmed by analyzing the events in the spills without the need of an external signal. An algorithm for the calculation of the period of the beam microstructure is developed. Small differences in the period between the spills impose the separate analysis for every spill. The period is calculated with 4 digits precision in nanosecond time scale, making a significant difference to the representation of the microbunch. In the end, the firmware related to the algorithm for the calculation of the period of the beam microstructure is developed using only the events in the spills. The simulation results show that it is possible the algorithm to be implemented in an FPGA and provide information about the period of the beam microstructure in real time. Moreover, a coincidence sorter is developed in order to provide real time coincidence detection. The simulation results for the two different architectures of the sorter that uses comparators with two and three inputs, are presented. The 3D spatial distribution and the 1D activity profiles of the coincidence events are constructed for inter-spill and in-spill data. The strong radiation background is visible in the reconstructed images, especially before the entrance surface of the phantom and at the end of the activity range with a tail. After filtering out the in-spill data by discarding the coincidence events that occur in a sub-interval of the microbunch, it is shown that the reconstructed image improves severely. In the 1D activity profile, one can observe that the number of coincidence events before the entrance surface of the phantom decreases significantly. This might happen because neutrons are discarded since they are detected a few ns later after the interaction of the beam with the nuclei. Results show that the signal to noise ratio (SNR), defined as the activity peak in the phantom divided by the background level, is improved by a factor of about 4.8 with respect to the in-spill signal. In the end, it is important to mention that this activity has been developed within the projects INSIDE and INFIERI (FP7-PEOPLE-2012-ITN project number 317446) funded by MIUR and EU respectively.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/11365/1013502
Attenzione
Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo