20.500.12556/RUL-183924 Optimizacija poti obsevanja v protonski terapiji z neprekinjenim obsevanjem Scan path optimization in proton therapy with continuous scanning Protonska terapija z neprekinjenim obsevanjem je tehnika radioterapije za zdravljenje raka, ki je hitrejša od klasične tehnike diskretnega obsevanja, saj žarka na poti od točke do točke ne ugašamo. Ker žarek med premikanjem dovaja dodatno, nenačrtovano dozo, moramo skupni čas potovanja skozi vse točke minimizirati. Algoritmi za reševanje problema trgovskega potnika, ki je danemu problemu zelo soroden, so dobro optimizirani, a prepočasni. Pri iskanju hitrejšega algoritma izkoristimo specifično postavitev točk v dvodimenzionalnih plasteh, dopuščamo pa tudi suboptimalne rešitve. V magistrskem delu analiziramo optimalne rešitve izbranih plasti ter teoretično in eksperimentalno ovrednotimo različne metode za optimizacijo poti obsevanja: privzeto serpentinsko metodo, požrešne metode, simulirano ohlajanje in metode, kjer si pomagamo z gručenjem točk. Najboljše rešitve vrne kombinirana metoda, kjer točke najprej združimo v gruče vzdolž hitrejše osi, nato pa gruče povežemo s simuliranim ohlajanjem. Ta kombinacija v delčku sekunde vrne rešitve, ki so od optimalnih slabše za nekaj odstotkov. Ob analizi rezultatov izpostavimo še praktične vidike optimizacije ter podamo nekaj predlogov za izboljšave in nadaljnje delo. Proton therapy with continuous scanning is a radiotherapy technique for cancer treatment that is faster than the conventional discrete scanning approach, as the beam remains on while moving between spots. Since the beam deposits additional, unplanned dose to the tissue during this movement, the goal is to minimize the beam's cumulative travel time across all spots. Algorithms for the travelling salesman problem, which is closely related to our problem, are well optimized but too slow. In the search for a faster algorithm, we exploit the specific geometric structure of spot placements in two-dimensional layers and also accept suboptimal solutions. We analyze optimal solutions for selected layers as well as theoretically and experimentally evaluate several scan path optimization methods: default serpentine method, greedy methods, simulated annealing, and clustering-based methods. The best results are produced by a combined method, which first clusters spots along the faster axis and then connects the clusters using simulated annealing. This combination returns solutions within a fraction of a second that are only a few percentage points worse than the optimum. While discussing the results, we also highlight practical considerations in scan path optimization and suggest ideas for future work. optimizacija problem trgovskega potnika radioterapija optimization travelling salesman problem radiotherapy true false false Slovenski jezik Angleški jezik Magistrsko delo/naloga 2026-06-21 08:15:37 2026-06-21 08:15:48 2026-06-22 09:00:05 0000-00-00 00:00:00 2026 0 0 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 1970-01-01 160493 282364931 21973.pdf 21973.pdf 1 68293A94C7F6592C2ACE5024E067A80B 0cf1d93f6858e7591f915b82921feb288d50b7b73877cecbbe9fc09577bd2b51 94b0e63b-6d38-11f1-9b0d-0050569b8976 https://repozitorij.uni-lj.si/Dokument.php?lang=slv&id=236236 Fakulteta za matematiko in fiziko 0 0 0