PET: Iterativna rekonstrukcija slike

Pozitronska emisijska tomografija (PET) je pomembna metoda medicinskega slikanja, ki temelji na zaznavi sevanja radioaktivnih \(\beta^{+}\) sevalcev. Te v majhnih količinah vnesemo v pacientovo telo kot del molekul t. i. radiofarmaka. V \(\beta^{+}\) razpadu nastane pozitron, ki se v snovi hitro ustavi in anihilira z bližnjim elektronom, energija pa se sprosti v obliki dveh antikolinearnih fotonov. Prestreže ju detektorski obroč, sestavljen iz scintilatorjev, ki absorbirajo ionizirajoče sevanje in ga pretvorijo v vidno svetlobo. To pa lahko zaznajo fotodetektorji. Iz meritev hkratnih fotonov je mogoče rekonstruirati 3D-porazdelitev radiofarmaka v telesu. Metoda za rekonstrukcijo meritve iz zadetkov, ki se trenutno najpogosteje uporablja, je iterativna rekonstrukcija, ki prek funkcije verjetja upošteva statistični značaj radioaktivnega razpada, kot tudi merilnega postopka. Pri tem se predpostavi linearna zveza med sliko in meritvijo, ki jo opisuje sistemska matrika. Rekonstrukcijski algoritem postopoma posodablja vrednosti slike, pri čemer se premika proti najbolj verjetni sliki (glede na konkretno meritev). Ločljivost je omejena z nekaterimi dejavniki (npr. doseg pozitrona v snovi, nekolinearnost anihilacije), izboljšuje pa se z razvojem natančnejših detektorjev. Razvijajo se tudi hitri detektorji, ki se uporabljajo v TOF PET, kjer se upošteva časovno razliko med detekcijo dveh fotonov. Kot alternativa iterativni rekonstrukciji se pojavljajo tudi metode, temelječe na konvolucijskih nevronskih mrežah, ki pa v klinični rabi še niso splošno sprejete.

PET: Iterative reconstruction

Positron emission tomography (PET) is an important medical imaging technique, based on the detection of radiation from radioactive \(beta^{+}\) emitters. These are introduced into the patient’s body in small amounts as part of molecules called radiopharmaceuticals. In \(\beta^{+}\) decay, a positron is produced, which quickly stops in the tissue and annihilates with a nearby electron, releasing energy in the form of two anti-collinear photons. These are intercepted by a detector ring, composed of scintillators that absorb ionizing radiation and convert it into visible light, which can then be detected by photodetectors. From the measurements of coincident photons, reconstruction of a 3D distribution of the radiopharmaceutical in the body is possible. Currently, prevalent reconstruction method is iterative reconstruction, which uses a likelihood function to account for the statistical nature of radioactive decay and measurement procedure. An assumption about a linear relationship between the image vector and the measurement vector is made, described by a system matrix. The reconstruction algorithm iteratively updates the image values, moving towards the most likely image (given the specific measurement). The resolution is limited by several factors (e.g., positron range in tissue, non-collinearity of annihilation), but it is improving with the development of more precise detectors. Fast detectors are also being developed for TOF PET, which takes into account the time difference between the detection of two photons. As an alternative to iterative reconstruction, convolutional neural network-based methods are emerging, but they lack validation for common clinical use.