First Advisor

Christof Teuscher

Term of Graduation

Summer 2021

Date of Publication


Document Type


Degree Name

Master of Science (M.S.) in Computer Science


Computer Science




Neural networks (Computer science), Gamma rays, Nuclides, Radiation



Physical Description

1 online resource (xv, 90 pages)


Radionuclide detection and identification are important tasks for deterring a potentially catastrophic nuclear event. Due to high levels of background radiation from both terrestrial and extraterrestrial sources, some form of noise reduction pre-processing is required for a gamma-ray spectrum prior to being analyzed by an identification algorithm so as to determine the identity of anomalous sources. This research focuses on the use of neuromorphic algorithms for the purpose of developing low power, accurate radionuclide identification devices that can filter out non-anomalous background radiation and other artifacts created by gamma-ray detector measurement equipment, along with identifying clandestine, radioactive material.

A sparse coding optimization solver, the Simple Spiking Locally Competitive Algorithm, is investigated and simulated for the tasks of radionuclide detection and identification. A convolutional neural network is used to filter the input signal to the identification algorithm to remove background radiation and detector noise. Both algorithms are designed to be neuromorphic, implemented in hardware using memristive devices, thus significantly reducing their necessary power consumption compared to software implementations.

The radionuclide identification algorithm is compared to Gamma Detector Response and Analysis Software, an industry standard package that is developed by Sandia National Laboratories. Our neuromorphic algorithm achieves a 91% accuracy with a high resolution detector and an 89% accuracy with a low resolution detector on the corresponding measured gamma-ray spectra test sets, both less than 2% below the benchmark, state of the art algorithm's performance on the same spectra. To determine the efficacy of using a neural network for background and noise reduction, identification results are compared between gamma-ray spectra with no noise reduction, the traditional standard of background subtraction, and using the presented convolutional neural network for denoising. Finally, the power consumption of the proposed neuromorphic algorithms is estimated and compared to the empirically determined power consumption of the Gamma Detector Response and Analysis Software, showing that they can achieve the same task with over a 99% reduction in power.


In Copyright. URI: This Item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you need to obtain permission from the rights-holder(s).

Persistent Identifier