James Biggs
About
My research project
Evaluation of scintillator detector performance using machine learningThe aim of this project is to investigate and evaluate the performance of scintillator-based radiation detectors in complex, real-world environments, where systems typically face trade-offs between resolution and detection efficiency. The research will look to compare traditional isotope identification methods, such as peak fitting algorithms, with a variety of machine learning based models.
The study will first determine the minimum useful detection threshold achievable by scintillator detectors under idealised conditions. It will then explore whether machine learning approaches can lower this threshold by extracting meaningful signals from noisy or count starved spectra. Building on this, the project will examine how artificial intelligence techniques can enhance detection capability in urban or cluttered environments, where background radiation, shielding, and environmental factors complicate source identification.
Using the recently expanded SIGMA 2 trial dataset, the work will apply analysis with real-world data to characterise detector responses and evaluate performance across realistic scenarios. Such work can be used to inform future detector deployment campaigns, as well as the methods used to identify isotopes present in spectra with a low signal-to-noise ratio.
Supervisors
The aim of this project is to investigate and evaluate the performance of scintillator-based radiation detectors in complex, real-world environments, where systems typically face trade-offs between resolution and detection efficiency. The research will look to compare traditional isotope identification methods, such as peak fitting algorithms, with a variety of machine learning based models.
The study will first determine the minimum useful detection threshold achievable by scintillator detectors under idealised conditions. It will then explore whether machine learning approaches can lower this threshold by extracting meaningful signals from noisy or count starved spectra. Building on this, the project will examine how artificial intelligence techniques can enhance detection capability in urban or cluttered environments, where background radiation, shielding, and environmental factors complicate source identification.
Using the recently expanded SIGMA 2 trial dataset, the work will apply analysis with real-world data to characterise detector responses and evaluate performance across realistic scenarios. Such work can be used to inform future detector deployment campaigns, as well as the methods used to identify isotopes present in spectra with a low signal-to-noise ratio.