Scott D. Metzler
VerifiedUniversity of Pennsylvania · Rehabilitation Medicine
Active 1991–2026
About
Scott D. Metzler, PhD, is a Research Professor of Radiology at the University of Pennsylvania's Perelman School of Medicine. His expertise lies in the development, characterization, and application of instrumentation for emission tomography, with a particular focus on collimation for single photon emission computed tomography (SPECT). He works on high-resolution cardiac imaging and has developed detailed models of pinhole-aperture penetration by high energy gamma rays, which is a limiting factor in high-resolution applications. Additionally, he has created sensitivity and resolution models for slit-slat collimation, a hybrid technique combining pinhole and parallel-beam collimation suitable for small-organ imaging in clinical settings. In his laboratory, he employs helical pinhole SPECT to acquire and reconstruct artifact-free images for whole-body small-animal applications, contributing to advancements in medical imaging technology.
Research topics
- Physics
- Optics
- Computer science
- Nuclear medicine
- Medicine
Selected publications
Single-Stage Lesion Identification in $^{68}$Ga-DOTATATE PET Images
IEEE Transactions on Biomedical Engineering · 2026-01-01
articleOBJECTIVE: Positron emission tomography (PET) is a commonly used imaging modality for assessment of neuroendocrine tumors (NETs), and lesion identification in PET images is a key step in the development of effective treatments. Deep neural networks have recently produced encouraging performance of automated lesion identification with PET imaging. However, most methods require a predefined region/volume of interest (ROI/VOI) or rely on a multi-stage, cascaded modeling pipeline, which often leads to low efficiency and/or high variability. In this paper, we propose a novel single-stage PET lesion detection method that does not need precomputed ROIs/VOIs, cascaded models or multimodal data. METHODS: We introduce a novel three-dimensional dual-decoder neural network, which contains a cross-decoder attention module to take as input gating signals from an auxiliary organ segmentation decoder and suppress irrelevant feature responses in the primary decoder of lesion detection. Additionally, we design and insert a new patchwise contrastive learning module into the primary decoder to enhance the network's discriminative power for lesions with varying volumes and shapes. RESULTS: We evaluate the proposed lesion identification method using multiple hepatic NET $^{68}$Ga-DOTATATE PET image datasets that are acquired from two different scanners. The method produces superior performance compared with the reference baseline and recent state-of-the-art approaches. CONCLUSION: We propose a novel single-stage framework, where both the cross-decoder attention and the patchwise contrastive learning are beneficial to improvement of lesion identification performance in PET images. SIGNIFICANCE: The proposed study has the potential to significantly improve the efficiency of clinical interpretation of PET imaging data.
Physics in Medicine and Biology · 2026-04-21
articleOpen accessSenior authorAbstract Objective 
Dynamic imaging in emission computed tomography allows the determination of physiological parameters that have important clinical implications. In myocardial imaging, three parameters of interest are the uptake and washout rates, K 1 and k 2 , respectively, and the blood-volume fraction, V. Theoretically, these parameters can be determined from time-activity curves (TACs) of the blood pool and the myocardium from injection and uptake of a perfusion tracer, if the matrix describing the relationship between the true TACs and the observed TACs is known. Herein, we numerically confirm these theoretical predictions and study the impact of experimental noise and matrix element uncertainty on these three parameters.

 Approach 
Using a one-compartment model, we simulated TACs for different values of K 1 , k 2 , and V . A mixing matrix involving recovery coefficients and spillover fractions was used to describe the transformation of the true concentration to the observed concentrations. Kinetic model fitting was employed to measure the three physical parameters, and we measured the bias of these fit results with respect to the true simulated values. We investigated a range of mixing matrices, rates of uptake and washout, and blood volume fractions.

 Main results 
We confirmed the theoretical prediction that K 1 , k 2 , and V can be simultaneously measured accurately if and only if all elements of the mixing matrix are known. There was an approximately linear effect of uncertainty in the measurement of the three physical parameters to the noise level on the TACs. Consistent with previous results, greater mixing via the mixing matrix or the blood-volume fraction resulted in greater uncertainty.

 Significance 
This work shows we can accurately measure the physical parameters like K 1 while also quantifying the uncertainty in their measurement due to noise on the TACs and mixing-matrix uncertainty.
IEEE Transactions on Radiation and Plasma Medical Sciences · 2025-11-24
articleThis study presents an experimental evaluation of the Dynamic Extremity SPECT (DE-SPECT) system, specifically engineered for precise, regionselective gamma-ray spectroscopy in the diagnosis of Peripheral Vascular Disease (PVD) in lower extremities. The system incorporates Cadmium Zinc Telluride (CZT) imaging spectrometers and dynamic dual-field-of-view (FOV) collimators to facilitate comprehensive, multifunctional molecular imaging. The CZT detectors, with Depth of Interaction (DOI) capabilities, deliver an exceptional energy performance across a wide energy range up to 600 keV. The novel dual-FOV aperture system allows selective imaging with two configurations: a 28-cm diameter wide FOV suitable for dual-leg or scout imaging and a 16-cm diameter high-resolution, and high-sensitivity (HR-HS) FOV designed for single-leg or focused imaging. Utilizing uniform phantoms, resolution phantoms, and multi-tracer phantoms, we experimentally assessed the system’s sensitivity, spatial resolution, and multi-tracer imaging capabilities. Spatial resolutions were approximately 6 mm in HR-HS-FOV mode and between 8 mm to 10 mm in wide-FOV mode. Peak-to-Valley ratios, indicative of image clarity, improved with enhanced DOI resolutions, rising from 1.03 to 1.22. The system’s ability to perform multi-tracer imaging, essential for deriving multifunctional molecular information, further highlights its potential to significantly enhance diagnostic accuracy for PVD.
2025-11-01
articleHigh-sensitivity, high-resolution brain-dedicated imaging systems have become critically important for advancing our understanding of complex neurological disorders such as Parkinson's disease, epilepsy, and Alzheimer's disease. In this paper, we present the initial experimental results of two hardware-enabled sampling enhancement strategies to be incorporated with our brain Single Photon Emission Computed Tomography (SPECT) scanner. The system integrates 320 Cadmium Zinc Telluride (CZT) detectors, configured in a compact cylindrical geometry. It features a dual field-of-view (FOV) design, offering both a 20-cm whole-brain imaging mode and a 5-cm high-resolution microscopic mode. For full-brain coverage, detectors are coupled with conventional 2-mm diameter pinhole collimators. In contrast, the microscopic imaging mode utilizes an innovative micro-ring aperture design with a 0.25-mm annular opening, enabling ultrahigh intrinsic spatial resolution while maintaining clinically viable sensitivity. A key feature of the system is its ability to enhance spatial resolution by incorporating non-redundant sampling positions. The collimator apertures are mounted on a precision linear motor platform, allowing fine translational shifts that generate sub-pixel projection variations. The complete collimator ring is mounted on a rotational belt, allowing for micro-rotations that generate angular sampling variations. Preliminary simulation results before sampling enhancement demonstrate a resolution of 6 mm for the larger FOV mode and 1 mm for the high-resolution mode using the micro-ring aperture configuration.
2025-11-01
articleWe present the DE-SPECT system, a region-selective clinical SPECT system designed for highperformance cardiovascular imaging. The system features six stationary CZT detector panels with dual-FOV collimators that enable rapid switching between a 28 cm wide field-of-view (FOV) and a 16 cm high-resolution, high-sensitivity (HR-HS) FOV. We evaluated DE-SPECT in large animal models of acute and chronic cardiovascular injury. Multi-isotope imaging with Tc-99m-tetrofosmin, I-123-MIBG, Ga-67 Citrate was used to visualize perfusion, sympathetic innervation and inflammation. A joint reconstruction algorithm was applied to combine a wide FOV with high resolution in the targeted region. We further demonstrated 1 -second dynamic imaging of tracer kinetics enabled by the system's high sensitivity and stationary geometry. These results establish the DE-SPECT as a versatile platform for dynamic, high-resolution, and multi-tracer imaging in cardiovascular disease.
Energy Scale-Factor Estimation for Use in Photomultiplier Tube Energy Calibration Using C-SPECT
IEEE Transactions on Radiation and Plasma Medical Sciences · 2025-05-01
articleOpen accessSenior authorAn array of photomultiplier tubes (PMTs) provides energy readout for gamma cameras, leading to event selection and positioning. However, operational and environmental changes, such as temperature, can cause PMTs to “drift” away from their nominal energy readouts and, therefore, require a correction procedure to return to their reference energies. We present two methods for determining the energy-scale change of each PMT using data collected on C-SPECT, a dedicated cardiac single-photon emission computational tomography (SPECT) scanner. A scan of a vertical line source of 99mTc provides the data from which we produce an energy histogram for each of the 130 PMTs. Each energy histogram is composed of events passing an energy-fraction selection to give events closest to the PMT center. We consider energy fractions ranging from 0.25% to 5.00%. For our analysis, we use bootstrapping to create data realizations as well as emulating energy-scale changes (simultaneously and independently for all PMTs) in the data. Using the average relative error as a measurement of the accuracy and the standard deviation from taking bootstrapped replicates of our data as a measurement of our precision, we determine the energy-scaling to within -0.05% <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\pm ~0.03$ </tex-math></inline-formula>% (mean and standard deviation, respectively).
2024-09-25
articleThe Dynamic Extremity SPECT (DE-SPECT) system employs state-of-the-art technology for precise, region-selective gamma-ray spectroscopy using Cadmium Zinc Telluride (CZT) imaging spectrometers and dynamic dual-field-of-view (FOV) collimators. This clinical system is designed to enhance the diagnostic accuracy of Peripheral Vascular Disease (PVD) assessments in the lower extremities by providing comprehensive, multifunctional molecular imaging. The CZT detectors used in the system offer an excellent energy performance (2.6 keV FWHM at 220 keV, 3.3 keV at 440 keV) spanning a wide energy range of up to 600 keV. The uniquely designed dual-FOV aperture system enables region-selective imaging capabilities with two configurations: a 28-cm diameter FOV for dual-leg imaging and a 16-cm diameter FOV for focused, high-resolution, and high-sensitivity imaging. The system’s unmatched capability facilitates in vivo simultaneous multi-tracer theragnostics within user-selected targeted regions. We have fully assembled the system and conducted phantom studies using a series of uniform phantoms, custom-made image quality (IQ) phantoms, and a resolution phantom filled with multiple radiotracers (Tc-99m, I-123, In-111, etc.) to evaluate the imaging performance of the DE-SPECT system.
Bioengineering · 2024-02-27 · 2 citations
articleOpen accessDeep learning (DL) algorithms used for DOTATATE PET lesion detection typically require large, well-annotated training datasets. These are difficult to obtain due to low incidence of gastroenteropancreatic neuroendocrine tumors (GEP-NETs) and the high cost of manual annotation. Furthermore, networks trained and tested with data acquired from site specific PET/CT instrumentation, acquisition and processing protocols have reduced performance when tested with offsite data. This lack of generalizability requires even larger, more diverse training datasets. The objective of this study is to investigate the feasibility of improving DL algorithm performance by better matching the background noise in training datasets to higher noise, out-of-domain testing datasets. 68Ga-DOTATATE PET/CT datasets were obtained from two scanners: Scanner1, a state-of-the-art digital PET/CT (GE DMI PET/CT; n = 83 subjects), and Scanner2, an older-generation analog PET/CT (GE STE; n = 123 subjects). Set1, the data set from Scanner1, was reconstructed with standard clinical parameters (5 min; Q.Clear) and list-mode reconstructions (VPFXS 2, 3, 4, and 5-min). Set2, data from Scanner2 representing out-of-domain clinical scans, used standard iterative reconstruction (5 min; OSEM). A deep neural network was trained with each dataset: Network1 for Scanner1 and Network2 for Scanner2. DL performance (Network1) was tested with out-of-domain test data (Set2). To evaluate the effect of training sample size, we tested DL model performance using a fraction (25%, 50% and 75%) of Set1 for training. Scanner1, list-mode 2-min reconstructed data demonstrated the most similar noise level compared that of Set2, resulting in the best performance (F1 = 0.713). This was not significantly different compared to the highest performance, upper-bound limit using in-domain training for Network2 (F1 = 0.755; p-value = 0.103). Regarding sample size, the F1 score significantly increased from 25% training data (F1 = 0.478) to 100% training data (F1 = 0.713; p < 0.001). List-mode data from modern PET scanners can be reconstructed to better match the noise properties of older scanners. Using existing data and their associated annotations dramatically reduces the cost and effort in generating these datasets and significantly improves the performance of existing DL algorithms. List-mode reconstructions can provide an efficient, low-cost method to improve DL algorithm generalizability.
Preprints.org · 2024-02-20 · 2 citations
preprintOpen accessDeep learning (DL) algorithms used for DOTATATE PET lesion detection typically require large, well-annotated training datasets. These are difficult to obtain due to low incidence of gastroenteopancreatic neuroendocrine tumors (GEP-NETs), and the high cost of manual annotation. Furthermore, networks trained and tested with data acquired from site specific PET/CT instrumentation, acquisition and processing protocols have reduced performance when tested with offsite data. This lack of generalizability requires even larger, more diverse training datasets. The objective of this study is to investigate the feasibility of improving DL algorithm performance by better matching the background noise in training datasets to higher noise, out-of-domain testing datasets. 68Ga-DOTATATE PET/CT datasets were obtained from two scanners: Scanner1, a state-of-the-art digital PET/CT (GE DMI PET/CT; n=83 subjects), and Scanner2, an older-generation analog PET/CT (GE STE; n=123 subjects). Set1, the data set from Scanner1 was reconstructed with standard clinical parameters (5 minutes; Q.Clear) and list-mode reconstructions (VPFXS 2, 3, 4, and 5-minutes). Set2, data from Scanner2, representing out-of-domain clinical scans, used standard iterative reconstruction (5 minutes; OSEM). Reconstructed datasets were divided into training, validation, and testing datasets in a 60%, 20%, 20% proportion, and a deep neural network was trained with each: Network1 for Scanner1 and Network2 for Scanner2. DL performance (Network1) was tested with out-of-domain test data (Set2). To evaluate the effect of training sample size, we tested DL model performance using a fraction (25%, 50%, and 75%) of Set1 for training. Scanner1, list-mode 2-minute reconstructed data demonstrated the most similar noise level compared that of Set2, resulting in the best performance (F1=0.713). This was not significantly different compared to the highest performance, upper-bound limit using in-domain training for Network2 (F1=0.755; p-value=0.103). Regarding sample size, the F1 score significantly increased from 25% training data (F1=0.478) to 100% training data (F1=0.713; p&amp;lt;0.001). List-mode data from modern PET scanners can be reconstructed to better match the noise properties of older scanners. Using existing data and their associated annotations dramatically reduces the cost and effort in generating these datasets, and significantly improves the performance of existing DL algorithms. List-mode reconstructions can provide an efficient, low-cost method to improve DL algorithm generalizability.
Abstract 4147473: Development of Pre-clinical Models of Coronary Microvascular Disease
Circulation · 2024-11-12
articleCoronary microvascular disease (CMVD), or disease of the coronary pre-arterioles, arterioles, and capillaries, accounts for 30-50% of ischemic heart disease. Progress in the field requires preclinical models to assess the coronary microvasculature. There are several risk factors for CMVD including age, metabolic syndrome, and hypercholesterolemia. Here, we evaluate the effect of these risk factors on coronary microvascular function in mice. Male and female C57BL/6 mice aged 12-42 weeks (n=29) were treated with 45% high fat diet (HFD) for six months or aged > 9 months. Apolipoprotein E knockout (ApoE-/-) was used to induce hypercholesterolemia as a second risk factor. To assess coronary microvascular function, we measured the intramyocardial blood volume (IMBV) under hyperemic (2.5% isoflurane) and basal (1.25% isoflurane) conditions, as previously reported. Briefly, we labeled red blood cells using pyrophosphate and Technetium 99m-pertechnatate and imaged the heart using µ-SPECT (MI Labs). Coronary microvascular function is reflected by the percent change in intramyocardial activity concentration between rest and stress conditions or △ IMBV. Outliers were removed based on Grubbs method (a=0.1) and groups were compared using Student's T test. p <0.05 was considered significant (Prism 9.3.1). Mice fed HFD had a 0.75-fold change in △ IMBV (23% ± 3.4 vs. 17% ± 2.4, p=0.03). Aging had a similar reduction in △ IMBV, 0.73-fold change (17% ± 2.7, p=0.01). We saw no sex-based difference in △ IMBV in our HFD and aged cohorts (18% vs 19%, 17% vs. 19%, p=0.79, p=0.74 , respectively), consistent with human cardiac perfusion PET imaging data. Exposure to two cardiovascular risk factors, ApoE-/- and HFD (0.48-fold change) or ApoE-/- and aging (0.54-fold change) additionally decreased △ IMBV (11% ± 2.3, p<0.01, 12% ± 2.6, p<0.01, respectively). Of note, ApoE-/- mice on a HFD had no evidence of obstructive epicardial coronary artery disease on histology. We evaluated coronary microvascular function in various mouse models using u-SPECT-based imaging methods and commercially available tracer. Exposure to cardiovascular risk factors induced CMVD, and the degree of CMVD was worse when two cardiovascular insults were used. Further studies using preclinical mouse models of CMVD may be helpful in assessing the effects of therapies and interventions.
Recent grants
Super-Resolution PET Using Stepping of a Deliberately Misaligned Bed
NIH · $433k · 2013–2016
Preclinical cardiac imaging package for clinical SPECT systems
NIH · $2.8M · 2012–2018
NIH · $1.4M · 2013
NIH · $954k · 2009
NIH · $793k · 2008
Frequent coauthors
- 53 shared
R.J. Jaszczak
- 32 shared
Roberto Accorsi
- 25 shared
K.L. Greer
Roger Williams Medical Center
- 24 shared
Steven C. Moore
- 24 shared
J.E. Bowsher
Duke Kunshan University
- 20 shared
Samuel Matej
University of Pennsylvania
- 17 shared
Yusheng Li
- 17 shared
Joel S. Karp
Institut National des Sciences Appliquées de Toulouse
- Resume-aware match score
- Save to shortlist
- AI-drafted outreach
See your match with Scott D. Metzler
PhdFit ranks faculty by your research interests, methods, and publications — grounded in their actual work, not templates.
- Free to start
- No credit card
- 30-second signup