Ocena prednosti rekonstrukcije pri TOF-PET s poznano časovno ločljivostjo dogodkov

Trinko, Urh (author), Studen, Andrej (mentor), Razdevšek, Gašper (comentor)

Pozitronska emisijska tomografija z meritvijo časa preleta (TOF-PET) je uveljavljena metoda funkcionalnega slikanja, pri kateri informacija o času preleta pomembno prispeva k izboljšanju kakovosti slike. Ključen korak metode je rekonstrukcija, kjer na podlagi zaznanih dogodkov, parov koincidenčnih fotonov, izluščimo porazdelitev radioaktivnega sledilca v pacientu. Učinek TOF je omejen s časovno ločljivostjo detektorskega sistema, ki se lahko razlikuje med posameznimi dogodki zaradi fizikalnih in tehničnih lastnosti detekcijske verige. Standardni rekonstrukcijski pristop te razlike zanemari in predpostavlja enako časovno nedoločenost za vse dogodke, opisano z Gaussovo porazdelitvijo. Ta predpostavka je precej oddaljena od realnosti, na primer v prihajajočih sistemih TOF-PET nove generacije, ki poleg scintilacijskih fotonov izkoriščajo tudi signal nastalih fotonov Čerenkova. V delu smo preverjali hipotezo, da lahko upoštevanje časovne ločljivosti v fazi rekonstrukcije izboljša klinično kakovost slik. V literaturi namreč primanjkuje študij, ki bi ta vpliv kvantitativno ovrednotile. Analiza temelji na Monte Carlo simulaciji slikanja različnih objektov, vključno s fantomom iz standarda NEMA NU-2, v PET-skenerju z geometrijo, podobno skenerju Siemens Biograph Vision. Simulacija je bila izvedena s programom GATE, rekonstrukcija slik pa s programom CASToR. Primerjali smo tri rekonstrukcijske pristope, ki se razlikujejo po obravnavi časovne nedoločenosti dogodkov, njihova učinkovitost pa je bila ovrednotena s standardnimi kazalniki. Rezultati kažejo, da upoštevanje razlik v časovni ločljivosti med dogodki v fazi rekonstrukcije izboljša kakovost slike, kadar je ta razlika dovolj velika. Ključno vlogo ima tudi absolutna vrednost časovne ločljivosti za dogodke z boljšo ločljivostjo: čim boljša je ta, tem bolj upravičen je preučevani pristop. Na primer, v sistemu, kjer polovica dogodkov dosega časovno ločljivost 50 ps, mora imeti druga polovica ločljivost vsaj ≥ 200 ps, da je upoštevanje razlik v rekonstrukciji še smiselno. Takšni rezultati so obetavni za detektorske sisteme v razvoju, kjer del dogodkov dosega zelo visoko časovno ločljivost.

• Language: Slovenian
• Keywords: TOF-PET, časovna ločljivost/časovno jedro, Monte Carlo simulacija, iterativna rekonstrukcija, kakovost slik
• Work type: Master's thesis/paper
• Typology: 2.09 - Master's Thesis
• Organization: FMF - Faculty of Mathematics and Physics
• Year: 2026
• PID: 20.500.12556/RUL-184956
• COBISS.SI-ID: 285968131
• Publication date in RUL: 18.07.2026

Secondary language

• Language: English
• Title: Benefits of time-resolution aware TOF-PET reconstruction

Abstract

Time-of-flight positron emission tomography (TOF-PET) is a well-established functional imaging modality in which time-of-flight information significantly improves image quality. During image reconstruction, the spatial distribution of a radioactive tracer is estimated from detected coincident photon pairs. The benefit of TOF information is limited by the timing resolution of the detector system, which may vary between individual events due to the physical and technical properties of the detection chain. The standard reconstruction approach neglects these variations and assumes the same Gaussian timing uncertainty for all events. This assumption may be inaccurate in emerging TOF-PET systems that exploit both scintillation and Cherenkov photons. In this thesis, we investigated the hypothesis that incorporating event-specific timing resolution into the reconstruction process can improve clinical image quality. Quantitative studies addressing this question remain scarce in the literature. The analysis is based on Monte Carlo simulations of several imaging scenarios, including the NEMA NU-2 image quality phantom, using a PET scanner geometry similar to that of Siemens Biograph Vision. Simulations were performed with GATE, while image reconstruction was carried out using CASToR. Three reconstruction approaches differing in their treatment of timing uncertainty were compared and evaluated using standard image quality metrics. The results show that accounting for differences in event timing resolution during reconstruction improves image quality when these differences are sufficiently large. The benefit also depends on the absolute timing resolution of the events with superior resolution: the better this resolution, the more advantageous the approach becomes. For example, in a system where half of the events achieve a timing resolution of 50 ps, the remaining half must have a timing resolution ≥ 200 ps for timing-resolution-aware reconstruction to remain beneficial. These findings are promising for next-generation detector technologies capable of achieving exceptionally high timing resolution for a subset of detected events.

• Keywords: TOF-PET, time resolution/timing kernel, Monte Carlo simulation, iterative reconstruction, image quality