Loading...
Stem effect correction using machine learning technique in radioluminescence sensors
Date
2024
Abstract
Stem Effect has been identified as a major characteristic of fiber optic-coupled dosimetry (FOCD) systems that limits the overall accuracy of the readings. The majority of work on compensating for the stem effect has revolved around sophisticated apparatus or analog signal processing techniques. In this paper, we present the results of a software-based technique, primarily relying on machine learning algorithms, to estimate and offset stem effects in FOCDs. We obtained data from a Ge-doped silica scintillator system, comprising a PMMA channel and PMT detector. Radiation was supplied by an Elekta Synergy radiotherapy machine, providing 6MV photon radiation at dose rates between 35 to 600 MU/min (1 MU = 1 cGy). The radioluminescence signal (RL) was acquired using two in-house assembled reader units, capable of capturing data at a 50µs gate time, with a buffer capacity of 1,000,000 points, effectively allowing a data acquisition window of 50 seconds. The radiation source operates with a pulse repetition frequency (prf) of 400Hz at the highest dose rate, delivering a pulse every 2.5 seconds. Therefore, the gate time of the data acquisition system offers the ability to capture the radiation in a pulse-by-pulse mode. The pulse-by-pulse mode, or time-resolved measurements, allows for analyzing and processing individual pulses for dose information or signal correction. By reading the response of nascent PMMA optical fiber channels, we may obtain the emission from the carrier fiber alone. Pulse-by-pulse measurements allow further capability to record the individual pulses comprising the stem effect. In this study, we first used the double carrier method to capture the raw data. Initial measurements were made using two nascent PMMA optical fibers. Once their response was standardized, the Ge-doped scintillator was attached to one of the PMMA fibers.
Supervisor
Description
Publisher
University of Limerick
Citation
20th Sensors and Their Applications Conference, 2024, Paper No: 56
Collections
Files
Loading...
Paper_56.pdf
Adobe PDF, 139.41 KB