System-on-chip-based radiation imaging system with pulse shape discrimination capability

Leland, John

Permalink

https://hdl.handle.net/2142/132816 Copy

Description

Title

System-on-chip-based radiation imaging system with pulse shape discrimination capability

Author(s)

Leland, John

Issue Date

2025-12-12

Director of Research (if dissertation) or Advisor (if thesis)

Di Fulvio, Angela

Committee Member(s)

Meng, Ling Jian

Department of Study

Nuclear, Plasma, & Rad Engr

Discipline

Nuclear, Plasma, Radiolgc Engr

Degree Granting Institution

University of Illinois Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

• stand-off imaging
• image reconstruction
• SiPM
• DOI
• depth-of-interaction
• pulse shape discrimination

Abstract

Compact scintillator array-based radiation imaging systems, capable of source characterization and localization of radioactive materials for nuclear security, safeguards, and non-proliferation applications, remain an unexplored technology as several technological challenges have prevented their adoption. These systems utilize large, dual-ended detector arrays for position sensitivity and advanced electronics capable of handling the high-channel readout. In this thesis, an analysis of the capability of each component of the imaging system is performed to determine its feasibility in performing spectroscopy, pulse shape discrimination (PSD), and image reconstruction, hence overcoming the constraint for deployment of hand-held scintillator-based stand-off imagers. A new PSD method, referred to as digital shaping (DS) PSD, was developed and optimized using the intrinsic shaping parameters of the commercially available Citiroc1A application-specific integrated circuit (ASIC) used for the high-channel SiPM readout. The optimization and analysis of its PSD capabilities showed that the ASIC is PSD capable with a figure of merit (FOM) of 0.78. This new method, unlike digitizer-based charge integration, does not require an external processing system and with increased flexibility in the on-board shaping parameters, the DS PSD performance can be further improved. To assess the scintillator performance for image reconstruction, a depth of interaction (DOI) study was performed to determine how the LYSO scintillator surface conditioning affected its position and energy resolution for image reconstruction. The fully polished crystals achieved the best energy resolution at 8.3% while the fully roughened crystal produced the lowest position uncertainties at 1.7mm and 2.9mm for the linear and sigmoid fits, respectively. Single-side roughening of the LYSO crystal provided a balance between the two performance metrics with an energy resolution of 10.7% and position uncertainties of 2.2mm and 3.0mm for the linear and sigmoid fits, respectively. An additional study was performed to determine if the same DOI results could be achieved using a simpler experimental setup. The new method utilized a collimated internal conversion source emitting mono-energetic electrons instead of relying on the traditional method of using the photons emitted from positron annihilation to geometrically collimate the beam size on the DOI crystal. For the polished LYSO crystal, the new method produced better linearity with an R2 of 0.984 compared to 0.968 from the traditional method. The new method also increased the coincidence count rate by five-times at 32.3 c/s . However, it also produced a poorer DOI uncertainty of 21.8 mm, nearly twice that of the traditional method. In conclusion, DOI can be best assessed using a positron emitting source in coincidence for a fine control of the position of interaction. Surface conditioning should preserve both energy resolution and DOI accuracy. When these aspects are considered collectively, a configuration encompassing a polished crystal with one side roughened with a diffuse reflector should be implemented. Finally, the FPGA-based readout provides a compact and programmable platform capable of on-board advanced analysis in real time. The maximum-likelihood expectation-maximization (MLEM) image reconstruction algorithms was implemented on a PYNQ-Z2 System-on-Chip board and optimized using High-Level Synthesis. The implementation showed the feasibility of using FPGAs for real-time, on-board pulse processing by comparing the reconstruction time on the board’s ARM Core and FPGA fabric. The optimized FPGA hardware implementation achieved a six-times speedup over the ARM Core with a reconstruction time of just 20.35 seconds. The final result was also compared to a Python ground truth, producing a negligible error of 0.001754. By integrating the findings of these studies, this thesis demonstrates the feasibility of a compact, dual-particle, high-channel imaging system capable of real-time source characterization, identification, and localization.

Graduation Semester

2025-12

Type of Resource

Thesis

Handle URL

https://hdl.handle.net/2142/132816

Copyright and License Information

Copyright 2025 John Leland

Owning Collections