Neural Network-Based Signal Processing for Enhanced Radiation Detection

Open Webpage

M. Kossakowski, S. Hauberg, I. Kuvvetli, S. Ringsborg Howalt Owe, G. Arvanitidis, A. Valverde Mahou, C. Budtz-Jorgensen

Published: 01 Jan 2025, Last Modified: 27 Jan 20262025 IEEE Nuclear Science Symposium (NSS), Medical Imaging Conference (MIC) and Room Temperature Semiconductor Detector Conference (RTSD)EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Integrating Artificial Intelligence (AI) and Machine Learning (ML) into signal processing can enhance high-resolution radiation detectors such as the 3D CZT drift strip detector at DTU Space, designed for space and medical applications. This study investigates Neural Network (NN) models for predicting radiation interaction positions, trained on synthetic data and evaluated with experimental data. A Feed-Forward Neural Network (FFNN) achieved comparable or improved positioning accuracy over conventional algorithms, notably near detector boundaries, aided by randomized synthetic electronic noise. Despite model and data limitations, the NN approach shows strong potential for AI-driven signal processing in space, healthcare, and security applications.