Matthew A. Kupinski

· Optical SciencesVerified

University of Arizona · Physics

Active 1995–2026

h-index32

Citations3.5k

Papers20341 last 5y

Funding$1.3M

Faculty page

See your match with Matthew A. Kupinski — sign in to PhdFit.Sign in

About

Matthew A. Kupinski is a faculty member in the Program in Applied Mathematics at the University of Arizona. His research interests include image science, image-quality assessment, system design, statistical decision theory, and nuclear-medicine imaging. His work encompasses X-ray imaging, near-infrared imaging, and radioactive transport theory. Kupinski is actively involved in research related to medical imaging and optical sciences, contributing to the development and assessment of imaging systems and techniques.

Research topics

Computer science
Artificial intelligence
Computer vision
Optics
Algorithm

Selected publications

UAMonteCarlo and MaterialProperties Swift/Metal Library
ReDATA · 2026-01-01
otherOpen access1st authorCorresponding
ReDATA curation note: source code files were submitted by the author but were removed from this dataset prior to publication due to licensing terms preventing public sharing. The description provided below as well as the included readme file may reference the removed files. Refer to README.txt for a list of items removed.UAMonteCarlo PackageUAMonteCarlo is a package that can model, x-ray/gamma-ray propagation as well as positron propagation through materials. This package requires the MaterialProperties package as well.MaterialPropertiesThis package computes x-ray and gamma-ray absorption, coherent scattering, incoherent scattering, pair production, and triplet production properties for elements, compounds, or mixtures of materials. This package should be compatible with all Swift platforms that support Foundation (Linux, macOS, iOS, and Windows).ElementsMaterialProperties supports the first 100 elements in the periodic table (Hydrogen through Fermium).An `Element` is an enumeration for the 100 elements. You likely don't need to use it directly. However, if you want to explore some properties of elements, you can, for example:``` let tungsten = Element.W print(tungsten.edges) // prints all the K, L, M, and N edges above 1keVprint(tungsten.naturalDensity) print(tungsten.name) // Tungstenprint(tungsten.symbol) // W print(tungsten.atomicWeight) // atomic weight in g/mole```To make Matter from an element you can``` let copper = Matter(element: .Cu)```which will utilize the natural density. You can override the natural density with``` let copper = Matter(element: .Cu, withDensity: 12.5)```CompoundsMaterialProperties can parse a chemical formula and return a material. Because a density cannot be inferred from the formula, you must provide a density. For example,``` guard let hydroxyApatite = Matter(compound: "Ca5(PO4)3(OH)", withDensity: 3.16) else { ... }guard let water = Matter(compound: "H2O", withDensity: 1.0) else { ... } ``` If the initializer returns `nil` then the program couldn't parse the chemical formula.Material databaseThis package include the entire Geant4 database of NIST materials along with their densities. To use, you can,``` guard let nai = Matter(fromDatabase: "Bone Compact Icru") else { ... }```you can override the density with the optional argument `Matter(fromDatabase: "Adipose Tissue Icrp", withDensity: 1.0)`. A list of all the materials supported is at the end of this readme.### Custom materialsYou can generate your own mixtures using``` public init(fractionsByWeight fromMatter: [(Matter, Double)], withDensity: Double)```and``` public init(fractionsByVolume fromMatter: [(Matter, Double)], withDensity: Double)```Both take an Tuple array of existing Matter and either the fraction by weight or volume of that Matter in the new material. The final density must specified currently. ### Generic initialization`Matter` includes an initializer from a `String` entry via``` public init?(name: String, withDensity: Double? = nil, descriptiveName: String? = nil) {```This will first search `name` in the database and if it does not find a match, will then try parsing the string as either a chemical formula or a single element. This is very useful for parsing user input and generating `Matter` and allows the user to not have to specify the flavor of material beforehand. ### Extracting propertiesYou can get the mass attenuation or linear attenuation coefficients for absorption, coherent scatter, incoherent scatter, pair production, and triplet production for energies ranging from 1keV to 100GeV. Spline fitting from the raw NIST XCOM data is used to generate this data. See the Unit Tests for validation. All provided energies are in keV, mass attenuation values are in g/cm², and linear attenuation coefficients are returned in 1/cm. NOTE: XCOM returns tables w/ MeV energies. ``` let material = Matter(element: .W)let energy = 140.5 // keV let energyArray = stride(from: 0, to: 8, by: 0.1).map { pow(10.0, $0) }let ρA = material.massAbsorption(at: energy) let ρAArray = material.massAbsorption(atEnergies: energyArray)let ρScC = material.massScatterCoherent(at: energy) let ρScI = material.massScatterIncoherent(at: energy)let ρPP = material.massPairProduction(at: energy) let ρTP = material.massTripletProduction(at: energy)let ρTotal = material.massAttenuation(at: energy, includeCoherent: false)for type in CoefficientType.allCases { print(material.massCoefficient(type, atEnergy: energy))}let µA = material.linearAbsorption(at: energy) let µScC = material.linearScatterCoherent(at: energy)let µScI = material.linearScatterIncoherent(at: energy) let µPP = material.linearPairProduction(at: energy)let µTP = material.linearTripletProduction(at: energy)let µTotalArray = material.linearAttenuation(atEnergies: energyArray) // default is to include coherent````materialProps` command-line programThis package includes a command-line executable targe called `materialProps` which will generate information about a provide matter string (element, compound, or database). It also provides linear attenuation in a grid or at a specific energy.``` USAGE: materialProps [--density ] [--atEnergy ] [--noTable] [--linear] [--noHeader] ARGUMENTS: Material string OPTIONS: --density Material density--atEnergy Only produce info at the following energies --noTable Hide the table of coefficients--linear Display linear attenuation instead of mass attenuation --noHeader Hide the header information-h, --help Show help information. ``` Example commands:``` materialProps "Ca5(PO4)3(OH)" --density 3.17 --atEnergy 140.5materialProps "Adipose Tissue Icrp" --noHeader materialProps W --density 17.7 --linearmaterialProps C```Final Notes- `Matter` and `Element` are both `Codable` so their information can be exported via JSON files (or any other format that supports `Codable`). - The GPU implementation of `Matter` is entirely contained in `UAMonteCarlo` and is not in this repository.- `Matter` is considered equal if it contains the same elements and weights, density, and atomic mass; the name is not considered when testing equality since `name` is a `get set` property. - Much of the information in this package is stored in custom JSON files. These files are only loaded when they are needed and then never again. There will be a slight overhead in the first use of these data structures.- `Matter` has an optional `atomicWeight` parameters which is present for elements and compounds but not for the database since mixtures don't have a single atomic weight associated with them. - The *natural density* of an element can be highly variable depending upon the source. Most of my data are simlar to Geant4's densities. However, there are a few small differences where I found answers on Wikipedia with more precise numbers. For the higher-Z elements, natural density is often "predicted".- Currently the coefficients for `.attenuation` computations include the coherent scatter in the sum. It is not obvious that this is correct for all applications.Material Database1,2-Dichlorobenzene 1,2-Dichloroethane5Cr-Mo-V_Stainless_Steel A-150 TissueA201.0_Aluminum A321_Stainless_SteelAF1410_High_Alloy_Steel AISI_1025_Carbon_SteelAZ91E_Magnesium AcetoneAcetylene AdenineAdipose Tissue Icrp AdiposeICRU44AdiposeTissue Air AlanineAluminum Oxide AmberAmmonia AnilineAnthracene B-100 BoneBGO BakeliteBarium Fluoride Barium SulfateBenzene Beryllium OxideBgo Blood IcrpBone Compact Icru Bone Cortical IcrpBone_Corticle Bone_ICRU_CompactBoron Carbide Boron OxideBrain Icrp ButaneC-552 C17200_TF00_Copper_BerylliumCZT Cadmium TellurideCadmium Tungstate Calcium CarbonateCalcium Fluoride Calcium OxideCalcium Sulfate Calcium TungstateCarbon Dioxide Carbon TetrachlorideCellulose Butyrate Cellulose CellophaneCellulose Nitrate Ceric SulfateCesium Fluoride Cesium IodideChlorobenzene ChloroformConcrete Custom_450_H900_Precipitation_Hardened_Stainless_SteelCyclohexane Dichlorodiethyl EtherDiethyl Ether Dimethyl SulfoxideEPDM_Non-Asbestos_Gasket_Material EthaneEthyl Alcohol Ethyl CelluloseEthylene Eye Lens IcrpFerric Oxide FerroborideFerrous Oxide Ferrous SulfateFreon-12 Freon-12B2Freon-13 Freon-13B1Freon-13I1 Gadolinium OxysulfideGallium Arsenide Gel Photo EmulsionGlandularFraction50 GlandularFraction90GlandularTissue Glass LeadGlass Plate GlutamineGlycerol GuanineGypsum ICRU_SoftTissueInconel_600 KaptonLYSO Lanthanum OxybromideLanthanum Oxysulfide Lead OxideLithium Amide Lithium CarbonateLithium Fluoride Lithium HydrideLithium Iodide Lithium OxideLithium Tetraborate LpropaneLung Icrp LungWithAirM3 Wax Magnesium CarbonateMagnesium Fluoride Magnesium OxideMagnesium Tetraborate Mercuric IodideMethane MethanolMix D Wax Ms20 TissueMuscle Skeletal Icrp Muscle Striated IcruMuscle With Sucrose Muscle Without SucroseMylar N,N-Dimethyl FormamideN-Butyl Alcohol N-HeptaneN-Hexane N-PentaneN-Propyl Alcohol NaI w/ ThalliumNaphthalene NitrileNitrobenzene Nitrous OxideNylon-11 Rilsan Nylon-6-10Nylon-6-6 Nylon-8062Nylon6 OctanePH13-8Mo_Stainless_Steel PH15-7Mo_Stainless_SteelPMMA ParaffinPhoto Emulsion Plastic Sc VinyltoluenePlexiglass Plutonium DioxidePolyacrylonitrile PolycarbonatePolychlorostyrene PolyethylenePolyoxymethylene PolypropylenePolystyrene PolytrifluorochloroethylenePolyvinyl Acetate Polyvinyl AlcoholPolyvinyl Butyral Polyvinyl ChloridePolyvinyl Pyrrolidone Polyvinylidene ChloridePolyvinylidene Fluoride Potassium IodidePotassium Oxide PropanePyrex Glass PyridineRubber Butyl Rubber NaturalRubber Neoprene Silicon DioxideSilver Bromide Silver ChlorideSilver Halides Silver IodideSkin Icrp Sodium CarbonateSodium Iodide Sodium MonoxideSod
Publisher DOI
Maximum-Likelihood--Based Position Decoding of Laser Processed Converging Pixel CsI: Tl Detectors for High-Resolution SPECT
arXiv (Cornell University) · 2026-02-09
articleOpen access
This study demonstrates the feasibility of a novel fabrication technique for high spatial resolution CsI: Tl scintillation detectors tailored for single photon emission computed tomography (SPECT) systems. Building upon our previously developed laser induced optical barrier (LIOB) method, which achieved high spatial resolution, excellent sensitivity, and 100% fabrication yield in CsI: Tl detectors, we extend this approach to a converging-pixel architecture. A CsI: Tl crystal array with converging pixels was designed and fabricated, featuring entrance-face pixels of 1.6x1.6 mm2 and photodetector side pixels of 2x2 mm2. To localize gamma-ray interactions, both the center of gravity (CoG) algorithm and a maximum-likelihood (ML) based decoding method were implemented. A custom built four axis motion platform was developed to deliver a finely collimated pencil beam at precisely controlled positions and angles across the array, enabling generation of a comprehensive dataset for prior knowledge and validation. The results demonstrate an energy resolution of 11.79+/-0.53% (collimated experiment) and a position localization accuracy of 1.00+/-0.42 mm (nearest neighbor interpolation), confirming that the proposed converging-pixel architecture, combined with statistical decoding algorithms, provides a promising path toward the development of high-performance SPECT detectors.
Publisher OA PDF
Maximum-Likelihood--Based Position Decoding of Laser Processed Converging Pixel CsI: Tl Detectors for High-Resolution SPECT
Open MIND · 2026-02-09
preprint
This study demonstrates the feasibility of a novel fabrication technique for high spatial resolution CsI: Tl scintillation detectors tailored for single photon emission computed tomography (SPECT) systems. Building upon our previously developed laser induced optical barrier (LIOB) method, which achieved high spatial resolution, excellent sensitivity, and 100% fabrication yield in CsI: Tl detectors, we extend this approach to a converging-pixel architecture. A CsI: Tl crystal array with converging pixels was designed and fabricated, featuring entrance-face pixels of 1.6x1.6 mm2 and photodetector side pixels of 2x2 mm2. To localize gamma-ray interactions, both the center of gravity (CoG) algorithm and a maximum-likelihood (ML) based decoding method were implemented. A custom built four axis motion platform was developed to deliver a finely collimated pencil beam at precisely controlled positions and angles across the array, enabling generation of a comprehensive dataset for prior knowledge and validation. The results demonstrate an energy resolution of 11.79+/-0.53% (collimated experiment) and a position localization accuracy of 1.00+/-0.42 mm (nearest neighbor interpolation), confirming that the proposed converging-pixel architecture, combined with statistical decoding algorithms, provides a promising path toward the development of high-performance SPECT detectors.
DOI
UAMonteCarlo and MaterialProperties Swift/Metal Library
ReDATA · 2026-01-01
otherOpen access1st authorCorresponding
ReDATA curation note: source code files were submitted by the author but were removed from this dataset prior to publication due to licensing terms preventing public sharing. The description provided below as well as the included readme file may reference the removed files. Refer to README.txt for a list of items removed.UAMonteCarlo PackageUAMonteCarlo is a package that can model, x-ray/gamma-ray propagation as well as positron propagation through materials. This package requires the MaterialProperties package as well.MaterialPropertiesThis package computes x-ray and gamma-ray absorption, coherent scattering, incoherent scattering, pair production, and triplet production properties for elements, compounds, or mixtures of materials. This package should be compatible with all Swift platforms that support Foundation (Linux, macOS, iOS, and Windows).ElementsMaterialProperties supports the first 100 elements in the periodic table (Hydrogen through Fermium).An `Element` is an enumeration for the 100 elements. You likely don't need to use it directly. However, if you want to explore some properties of elements, you can, for example:``` let tungsten = Element.W print(tungsten.edges) // prints all the K, L, M, and N edges above 1keVprint(tungsten.naturalDensity) print(tungsten.name) // Tungstenprint(tungsten.symbol) // W print(tungsten.atomicWeight) // atomic weight in g/mole```To make Matter from an element you can``` let copper = Matter(element: .Cu)```which will utilize the natural density. You can override the natural density with``` let copper = Matter(element: .Cu, withDensity: 12.5)```CompoundsMaterialProperties can parse a chemical formula and return a material. Because a density cannot be inferred from the formula, you must provide a density. For example,``` guard let hydroxyApatite = Matter(compound: "Ca5(PO4)3(OH)", withDensity: 3.16) else { ... }guard let water = Matter(compound: "H2O", withDensity: 1.0) else { ... } ``` If the initializer returns `nil` then the program couldn't parse the chemical formula.Material databaseThis package include the entire Geant4 database of NIST materials along with their densities. To use, you can,``` guard let nai = Matter(fromDatabase: "Bone Compact Icru") else { ... }```you can override the density with the optional argument `Matter(fromDatabase: "Adipose Tissue Icrp", withDensity: 1.0)`. A list of all the materials supported is at the end of this readme.### Custom materialsYou can generate your own mixtures using``` public init(fractionsByWeight fromMatter: [(Matter, Double)], withDensity: Double)```and``` public init(fractionsByVolume fromMatter: [(Matter, Double)], withDensity: Double)```Both take an Tuple array of existing Matter and either the fraction by weight or volume of that Matter in the new material. The final density must specified currently. ### Generic initialization`Matter` includes an initializer from a `String` entry via``` public init?(name: String, withDensity: Double? = nil, descriptiveName: String? = nil) {```This will first search `name` in the database and if it does not find a match, will then try parsing the string as either a chemical formula or a single element. This is very useful for parsing user input and generating `Matter` and allows the user to not have to specify the flavor of material beforehand. ### Extracting propertiesYou can get the mass attenuation or linear attenuation coefficients for absorption, coherent scatter, incoherent scatter, pair production, and triplet production for energies ranging from 1keV to 100GeV. Spline fitting from the raw NIST XCOM data is used to generate this data. See the Unit Tests for validation. All provided energies are in keV, mass attenuation values are in g/cm², and linear attenuation coefficients are returned in 1/cm. NOTE: XCOM returns tables w/ MeV energies. ``` let material = Matter(element: .W)let energy = 140.5 // keV let energyArray = stride(from: 0, to: 8, by: 0.1).map { pow(10.0, $0) }let ρA = material.massAbsorption(at: energy) let ρAArray = material.massAbsorption(atEnergies: energyArray)let ρScC = material.massScatterCoherent(at: energy) let ρScI = material.massScatterIncoherent(at: energy)let ρPP = material.massPairProduction(at: energy) let ρTP = material.massTripletProduction(at: energy)let ρTotal = material.massAttenuation(at: energy, includeCoherent: false)for type in CoefficientType.allCases { print(material.massCoefficient(type, atEnergy: energy))}let µA = material.linearAbsorption(at: energy) let µScC = material.linearScatterCoherent(at: energy)let µScI = material.linearScatterIncoherent(at: energy) let µPP = material.linearPairProduction(at: energy)let µTP = material.linearTripletProduction(at: energy)let µTotalArray = material.linearAttenuation(atEnergies: energyArray) // default is to include coherent````materialProps` command-line programThis package includes a command-line executable targe called `materialProps` which will generate information about a provide matter string (element, compound, or database). It also provides linear attenuation in a grid or at a specific energy.``` USAGE: materialProps [--density ] [--atEnergy ] [--noTable] [--linear] [--noHeader] ARGUMENTS: Material string OPTIONS: --density Material density--atEnergy Only produce info at the following energies --noTable Hide the table of coefficients--linear Display linear attenuation instead of mass attenuation --noHeader Hide the header information-h, --help Show help information. ``` Example commands:``` materialProps "Ca5(PO4)3(OH)" --density 3.17 --atEnergy 140.5materialProps "Adipose Tissue Icrp" --noHeader materialProps W --density 17.7 --linearmaterialProps C```Final Notes- `Matter` and `Element` are both `Codable` so their information can be exported via JSON files (or any other format that supports `Codable`). - The GPU implementation of `Matter` is entirely contained in `UAMonteCarlo` and is not in this repository.- `Matter` is considered equal if it contains the same elements and weights, density, and atomic mass; the name is not considered when testing equality since `name` is a `get set` property. - Much of the information in this package is stored in custom JSON files. These files are only loaded when they are needed and then never again. There will be a slight overhead in the first use of these data structures.- `Matter` has an optional `atomicWeight` parameters which is present for elements and compounds but not for the database since mixtures don't have a single atomic weight associated with them. - The *natural density* of an element can be highly variable depending upon the source. Most of my data are simlar to Geant4's densities. However, there are a few small differences where I found answers on Wikipedia with more precise numbers. For the higher-Z elements, natural density is often "predicted".- Currently the coefficients for `.attenuation` computations include the coherent scatter in the sum. It is not obvious that this is correct for all applications.Material Database1,2-Dichlorobenzene 1,2-Dichloroethane5Cr-Mo-V_Stainless_Steel A-150 TissueA201.0_Aluminum A321_Stainless_SteelAF1410_High_Alloy_Steel AISI_1025_Carbon_SteelAZ91E_Magnesium AcetoneAcetylene AdenineAdipose Tissue Icrp AdiposeICRU44AdiposeTissue Air AlanineAluminum Oxide AmberAmmonia AnilineAnthracene B-100 BoneBGO BakeliteBarium Fluoride Barium SulfateBenzene Beryllium OxideBgo Blood IcrpBone Compact Icru Bone Cortical IcrpBone_Corticle Bone_ICRU_CompactBoron Carbide Boron OxideBrain Icrp ButaneC-552 C17200_TF00_Copper_BerylliumCZT Cadmium TellurideCadmium Tungstate Calcium CarbonateCalcium Fluoride Calcium OxideCalcium Sulfate Calcium TungstateCarbon Dioxide Carbon TetrachlorideCellulose Butyrate Cellulose CellophaneCellulose Nitrate Ceric SulfateCesium Fluoride Cesium IodideChlorobenzene ChloroformConcrete Custom_450_H900_Precipitation_Hardened_Stainless_SteelCyclohexane Dichlorodiethyl EtherDiethyl Ether Dimethyl SulfoxideEPDM_Non-Asbestos_Gasket_Material EthaneEthyl Alcohol Ethyl CelluloseEthylene Eye Lens IcrpFerric Oxide FerroborideFerrous Oxide Ferrous SulfateFreon-12 Freon-12B2Freon-13 Freon-13B1Freon-13I1 Gadolinium OxysulfideGallium Arsenide Gel Photo EmulsionGlandularFraction50 GlandularFraction90GlandularTissue Glass LeadGlass Plate GlutamineGlycerol GuanineGypsum ICRU_SoftTissueInconel_600 KaptonLYSO Lanthanum OxybromideLanthanum Oxysulfide Lead OxideLithium Amide Lithium CarbonateLithium Fluoride Lithium HydrideLithium Iodide Lithium OxideLithium Tetraborate LpropaneLung Icrp LungWithAirM3 Wax Magnesium CarbonateMagnesium Fluoride Magnesium OxideMagnesium Tetraborate Mercuric IodideMethane MethanolMix D Wax Ms20 TissueMuscle Skeletal Icrp Muscle Striated IcruMuscle With Sucrose Muscle Without SucroseMylar N,N-Dimethyl FormamideN-Butyl Alcohol N-HeptaneN-Hexane N-PentaneN-Propyl Alcohol NaI w/ ThalliumNaphthalene NitrileNitrobenzene Nitrous OxideNylon-11 Rilsan Nylon-6-10Nylon-6-6 Nylon-8062Nylon6 OctanePH13-8Mo_Stainless_Steel PH15-7Mo_Stainless_SteelPMMA ParaffinPhoto Emulsion Plastic Sc VinyltoluenePlexiglass Plutonium DioxidePolyacrylonitrile PolycarbonatePolychlorostyrene PolyethylenePolyoxymethylene PolypropylenePolystyrene PolytrifluorochloroethylenePolyvinyl Acetate Polyvinyl AlcoholPolyvinyl Butyral Polyvinyl ChloridePolyvinyl Pyrrolidone Polyvinylidene ChloridePolyvinylidene Fluoride Potassium IodidePotassium Oxide PropanePyrex Glass PyridineRubber Butyl Rubber NaturalRubber Neoprene Silicon DioxideSilver Bromide Silver ChlorideSilver Halides Silver IodideSkin Icrp Sodium CarbonateSodium Iodide Sodium MonoxideSod
Publisher DOI
Simulation-Based Performance Evaluation of the DC-SPECT System
2025-11-01
article
This work evaluates the system performance of a dynamic cardiac SPECT (DC-SPECT) system using Monte Carlo simulations. We assessed the system's sensitivity across the field of view (FOV) and evaluated spatial resolution using reconstructed images of a Derenzo phantom, a Jaszczak phantom, and an XCAT cardiac phantom. Results show that the system achieves a resolution of 5.5 mm for hot rods without background, and 6.0 mm for cold rods with warm background. The sensitivity ranges from 5 to 9 times greater than conventional dual-head SPECT over a 15 cm diameter spherical FOV. This study demonstrates that the current design of DC-SPECT has the potential to support effective cardiac imaging.
Publisher DOI
Waveform Measurement of Mobility-Lifetime Products of CZT Detector Samples
2025-11-01
article
This paper presents a waveform measurement of the induced signals from the bias-induced drift of the charge carriers — electrons and holes created by charge particles and high-energy photon interactions — within single-pixel CZT detector samples. The detectors were made from the material grown using the Accelerated Crucible Rotation Technique. CZT is well-recognized as a promising semiconductor material for gamma detection in SPECT and PET imaging due to the advantages of high energy resolution and high stopping power. To fully use these advantages, detectors made from spectroscopic-grade CZT material and understanding its trapping properties are necessary. The trapping property of a material is usually described by mobility-lifetime <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mu \tau$</tex> products of charge carriers. In this work, we used waveform readout to measure the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mu \tau$</tex> products of electrons and holes of various samples and detect depth-dependent photon signals, which can be later used in the estimation of depth-of-interaction (DOI) and corrections of the energy spectrum. We described the waveform signal capture with a mathematical model and estimated <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mu_{\mathrm{e}} \tau_{\mathrm{e}}=1.19 * 10^{-3} ~\text{cm}^{2} / \mathrm{V}$</tex> and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mu_{\mathrm{h}} \tau_{\mathrm{h}}=4.18 * 10^{-5} ~\text{cm}^{2} / \mathrm{V}$</tex> for a selected sample. This work presents an alternative method for assessing the quality of CZT in high-energy photon detection in medical imaging systems.
Publisher DOI
Machine learning-assisted event classification in cadmium zinc telluride positron emission tomography detectors leveraging entanglement-informed angular correlations
Scientific Reports · 2025-12-20
articleOpen access
Gamma-positron imaging with tracers that emit a prompt γ (> 511 keV) is vulnerable to Compton down-scatter leaking into the 511-keV window and mimicking true annihilation pairs. Conventional Positron Emission Tomography (PET) systems reconstruct annihilation events without leveraging that the two 511-keV photons are not only orthogonally polarized but also produced in a Bell-entangled state. The polarization correlations of this entanglement imprint themselves in Compton scattering kinematics, particularly the relative azimuthal scattering angle ([Formula: see text]), offering a physics-informed handle for event discrimination. We present a machine-learning framework that exploits these quantum-encoded features to resolve true lines of response (LORs) and reject random coincidences in a dual-panel cadmium zinc telluride (CZT) system. Detected events were categorized into one-photoelectric (1P) and Compton (1C) interaction patterns, yielding four candidate interaction sequences per event. Each event was represented as a 4 × 21 feature matrix comprising spatial coordinates, energy deposits, and angular descriptors, including [Formula: see text] and polar scattering angle θ. Feature ablation with five-fold cross-validation revealed that the combination of energy and [Formula: see text] provided the highest discriminative power (Area Under the Receiver Operating Characteristic Curve (ROC-AUC) 0.87-0.95), followed by energy alone (ROC-AUC 0.85-0.95), while inclusion of spatial coordinates with energy and [Formula: see text] ranked third, achieving consistent performance across folds (ROC-AUC 0.81-0.91). These results demonstrate that incorporating entanglement-sensitive angular features into learning pipelines can suppress prompt contamination while preserving true LORs in a gamma-positron imaging system.
Publisher OA PDF DOI
Examination of aperture layout designs for an adaptive‐stationary multi‐pinhole brain‐dedicated SPECT system
Medical Physics · 2025-05-11 · 1 citations
article
BACKGROUND: Organ specific multi-pinhole (MPH) SPECT imaging could potentially improve the sensitivity/resolution trade-off and image quality (IQ), while facilitating the use of a variety of imaging-agents, thereby addressing diagnostic, quantitative, and research clinical needs. PURPOSE: Investigate through simulation six different MPH aperture-layout designs, plus variations in projection multiplexing (MUX) and truncation, for a prototype brain-dedicated MPH SPECT system, named AdaptiSPECT-C, to understand tradeoffs for such choices and guide selection of an optimal design for construction of the actual AdaptiSPECT-C system. METHODS: The prototype AdaptiSPECT-C system investigated herein employs 25 MPH gamma-camera modules arranged in three rings to image a 21 cm diameter spherical volume-of-interest (VOI). With a focal point (FP) to center of detector distances of 38.7 cm, the pinhole aperture diameters were constrained to provide a calculated spatial resolution of 8 mm at the FP. Variations in the number of pinhole (PH) apertures, FP to aperture distance, PH layout, temporal changes in MUX, and extent-of-truncation of the projection images were investigated. Designs of the aperture layouts were used to create inputs for GATE and analytic simulations of a sphere phantom with uniform Tc-99 m activity filling the VOI, to assess MUX, detector utilization, and uniformity in reconstructed slices. We investigated axial and angular sampling using customized-spherical Defrise and Derenzo phantoms. Finally, we assessed reconstructed IQ and activity quantification in reconstructions of analytic simulations of the XCAT digital anthropomorphic phantom with activity and attenuation distributions mimicking clinical-SPECT brain-perfusion imaging. For each phantom, comparison was also made to imaging with a dual-headed SPECT system with low-energy high-resolution (LEHR) parallel-hole (Vertex high resolution [VXHR]) collimators. RESULTS: Sensitivity at the FP (SENS) for a Tc-99 m source in air calculated relative to a clinical dual-headed SPECT system with VXHR collimators was 2.7x higher for a single aperture with no MUX or truncation, increased to 5.7x for five apertures with limited VOI truncation and MUX, and decreased to 2.5x with 13 apertures with limited MUX. For the spherical tub phantom, limited truncation did not impact uniformity, MUX decreased it, and temporal shuttering of projections helped lessen this impact. Visually, the 6.4 mm rods were generally well differentiated for the single central apertures. For designs with four or more apertures, all the 4.8 mm rods were well differentiated visually. Projection images of the XCAT phantom acquired for an imaging time that would result in the minimum clinically recommended count-level for brain perfusion imaging with parallel-hole collimators, showed low MUX of the brain structures for all of the MPH aperture layout designs. The best reconstructions for the XCAT phantom, both visually and quantitatively, were obtained with the design using 4- or 5-PH-apertures for the aperture-layout design that included MUX and some truncation of imaging. CONCLUSIONS: We determined for a prototype brain-dedicated MPH SPECT employing 25 camera modules in three rings with different PH layout designs imaging a 21 cm diameter spherical VOI, that a system with five apertures per module provided the best SENS, and IQ of the XCAT brain phantom, both visually and numerically.
Publisher DOI
Analytical methods for system matrix calculation and spatial resolution evaluation of DC-SPECT system
Physics in Medicine and Biology · 2025-07-10 · 3 citations
articleOpen accessCorresponding
Abstract Objective. To develop and evaluate an analytical method for calculating the system matrix of a dynamic cardiac single photon emission computed tomography (DC-SPECT) system, eliminating the need for computationally intensive Monte Carlo (MC) simulations. Approach. An analytical model was proposed for system matrix generation, incorporating solid angle calculations adapted for square-shaped pinhole collimators. The resulting sensitivity maps were validated against MC simulation results. Image reconstructions using the proposed analytical model were compared to those using an MC-simulated system matrix. Additionally, the overall system performance, including sensitivity and spatial resolution, was assessed using MC simulations. Main results. The analytical sensitivity map showed good agreement with MC-based results. Reconstructed images using the analytical model preserved key features but showed reduced performance compared to MC-based reconstructions. MC simulations of DC-SPECT demonstrated a spatial resolution of 5.5 mm for hot regions and 6.0 mm for cold regions, with an overall system sensitivity of 0.07% over a 15 cm diameter spherical field of view. Significance. The proposed analytical method offers a fast and practical alternative to MC-based system matrix generation, enabling efficient system design and prototyping. While its use in clinical image reconstruction requires further evaluation, the model provides a promising tool for accelerating the development of next-generation DC-SPECT systems.
Publisher OA PDF DOI
Development and Performance Evaluation of a Laser Processed CsI:Tl Detector with Converging Pixels
2025-11-01
article
Accurate and efficient detection of gamma photons is fundamental to the performance of nuclear imaging systems such as SPECT. The design and microstructural engineering of these detectors directly influence spatial resolution, energy resolution, and system sensitivity, making innovative fabrication techniques essential for next-generation imaging. Previously, we proposed laser induced optical barrier (LIOB) technique promising high spatial resolution and high sensitivity CsI:Tl detectors with 100% process yield. Building on this foundation, the present study extends the LIOB approach to a converging-pixel CsI crystal design tailored for SPECT applications. A <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$50 \times 50 \times 8.05 ~\text{mm}^{3}$</tex> thick <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{CsI}: \text{Tl}$</tex> scintillator is laser processed to <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$25 \times 25$</tex> pixels with 2 mm pixel pitch at the photodetector plane and 1.6 mm pitch at the entrance plane. The scintillator is backed by <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$8 \times 8$</tex> array of Hamamatsu S14161 with 6 mm pixels. A custom 4-axis motion platform was built to deliver a collimated pencil beam at precisely controlled positions and angles across the crystal array, providing the experimental dataset needed for algorithm training and validation. Preliminary decoding of interaction positions using the COG and deep learning algorithms was performed. The decoding results indicate that the deep learning method significantly outperforms the COG method. Overall, this study confirms that the converging-pixel architecture, when paired with advanced statistical or machine learning-based processing, represents a promising approach for developing high-performance SPECT detectors.
Publisher DOI

Recent grants

NIH Grant R01EB002146
NIH · $723k · 2008
NIH Grant RC1EB010974
NIH · $545k · 2012

Frequent coauthors

Eric Clarkson
University of Arizona
89 shared
Harrison H. Barrett
70 shared
Lars R. Furenlid
56 shared
Phillip H. Kuo
University of Arizona
27 shared
Abhinav K. Jha
Mallinckrodt (United States)
22 shared
Michael A. King
University of Massachusetts Chan Medical School
20 shared
R. Garrett Richards
University of Arizona
14 shared
Kevin A. Gross
Optics Technology (United States)
13 shared

Resume-aware match score
Save to shortlist
AI-drafted outreach

See your match with Matthew A. Kupinski

PhdFit ranks faculty by your research interests, methods, and publications — grounded in their actual work, not templates.

Join the waitlist How it works

Free to start
No credit card
30-second signup

Find professors who actually fit you