Efficient On-Board Image Reconstruction Algorithms for a High-Density-Channel Imager

Year
2024
Author(s)
John Leland - Department of Nuclear, Plasma, and Radiological Engineering, University of Illinois
Ming Fang - Department of Nuclear, Plasma, and Radiological Engineering, University of Illinois
Revanth Chappidi - Department of Electrical and Computer Engineering, University of Illinois
Cecilia Tarpau - School of Mathematical and Computer Sciences, Heriot-Watt University, Maxwell Institute for Mathematical Sciences, Bayes Center, University of Edinburgh
Yoann Altmann - School of Engineering and Physical Sciences, Heriot-Watt University
Angela Di Fulvio - Department of Nuclear, Plasma, and Radiological Engineering, University of Illinois
Abstract

In this paper, we have successfully demonstrated the ability to implement and optimize two image reconstruction algorithms, filtered back projection (FBP) and maximum-likelihood expectation maximization (MLEM), for stand-off gammaray imagereconstruction. The algorithms were implemented on an inexpensive and small Xilinx Pynq-Z2 evaluation board which hosts a Kintex-7 field programmable gate array (FPGA). FPGAs are customizable platforms with similar functionality to fast waveform digitizers while also offering the capability to perform on-board advanced analysis, in real time. These characteristics make FPGAs an ideal platform for the development of advanced radiation detection systems for rapid source characterization and localization. We implement and optimize the image reconstruction algorithms directly on the FPGA hardware. Our optimized solution achieved rapid source localization in 20.35 seconds, which is a 5.96× speedup over its non-optimized counterpart. With this system, we can determine the position of a 100 µCi 137Cs source placed 1 meter from the detector in under 30 seconds- 5 to 10 seconds of measurement time and 20 seconds to perform the reconstruction. The truncation error due to the reduced precision did not degrade the image contrast significantly.