About

Peter B. Noël, M.S., Ph.D., is an Associate Professor of Radiology at the University of Pennsylvania's Perelman School of Medicine. He serves as the Chief Technology Officer of One Penn Radiology, the Director of CT Research within the Department of Radiology, and Co-Director of the Center for Imaging Research with Radiation in Radiology (CIR³). He is also the Director of the Computed Tomography Center (CTC) and a member of the Radiation Research Safety Committee. Dr. Noël's research focuses on developing x-ray and Computed Tomography (CT) technology, with an emphasis on harmonizing hardware and software, reconstruction, and post-processing to create integrated clinical solutions. He has contributed significantly to the clinical translation of novel diagnostic imaging technologies, including photon-counting CT. His work also involves the development of advanced 3D printed CT phantoms, including patient-specific lung phantoms, which are available to the research community. His research extends to fundamental x-ray physics questions, particularly phase-contrast and dark-field imaging. With over 190 peer-reviewed publications and a patent portfolio, Dr. Noël is recognized for his contributions to the field of medical imaging.

Research topics

• Artificial Intelligence
• Computer Science
• Medical physics
• Medicine
• Mathematics
• Nuclear medicine
• Nanotechnology
• Materials science
• Statistics
• Radiology

Selected publications

• Author Correction: Combined inhibition of BET family proteins and histone deacetylases as a potential epigenetics-based therapy for pancreatic ductal adenocarcinoma

  Nature Medicine · 2026-01-29

  articleOpen access
  PublisherOA PDFDOI

• Abstract No. 337 Modeling Temperature Changes in a Cryoablation Phantom Using Spectral CT

  Journal of Vascular and Interventional Radiology · 2026-03-23

  articleOpen accessSenior author
  PublisherOA PDFDOI

• Deep Learning-Based Attenuation Correction in TOF-PET Using Histo-Image Data Partitioning and Dual-View Scout Images

  2025-11-01

  article

  Clinical PET scanners integrate CT for anatomical correlation and PET corrections, but there are scenarios that do not require the CT and may benefit from reduced radiation and avoidance of artefacts (quantification biases) due to motion between PET and CT scans. Deep learning (DL) offers a promising path toward accurate CT-less PET. Existing DL-based AC approaches rely on reconstructed non-attenuation corrected (NAC) PET images, which no longer contain raw time-offlight (TOF) data. To better exploit the directional and attenuation-related information embedded in TOF-PET, we propose a DL framework using multi-view histo-images, direct representations of TOF events in image space, as input instead of NAC images. We also explore the integration of scout images (2D radiographic projections acquired for scan planning), which offer some anatomical information at <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\sim 10$</tex> x lower dose than the low-dose CT, to potentially improve DL-based AC. All DL models performed reasonably well, with SUV biases within <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\pm 1$</tex> SUV unit (<10%). Leveraging anatomical information from dual-view scouts further reduced biases, enabling quantitative PET images without utilizing a low-dose CT. For stand-alone (dedicated) PET systems, the use of multi-view histo-images demonstrated comparable performance to previous methods utilizing NAC PET images as inputs, with the added advantage of faster, more efficient image generation.

  PublisherDOI

• Pediatric Sparse Spectral CT via Hybrid Photon-Counting and kVp-Switching

  2025-11-01

  articleSenior author

  While spectral CT has improved adult diagnostic imaging, it remains primarily optimized for larger body sizes, neglecting key pediatric imaging challenges-namely, small anatomical detail, low intrinsic contrast, and high radiation sensitivity. Photon-counting detectors (PCDs) offer high resolution at low doses, but even ideal PCDs suffer from excessive noise and bias caused by minimal counts in ultra-low dose pediatric imaging. Further, generating spectral results by splitting low counts into bins is not feasible. This work proposes a novel sparse spectral imaging paradigm tailored to pediatric needs, enabling low-dose, ultra-high resolution spectral CT. A PCD operated in high-resolution mode (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$0.5 \times 0.5 ~\text{mm}^{2}$</tex>) acquires most of the scan using a low 70 kVp tube voltage, summing photon counts across energy bins, functioning in non-spectral mode. During designated sparse intervals, the system transitions into a rapid kilovolt-peak switching (kVp-S) mode using 110 kVp to activate spectral imaging with the detector's energy-discriminating functionality using <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$1 \times 1 ~\text{mm}^{2}$</tex> pixel size. A Monte Carlo simulation was used to evaluate the proposed pediatric imaging protocol. At equivalent doses compared to a constant 100 kVp reference scan, the sparse spectral protocol produces <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$2 x$</tex> spectral SNR performance and 97 % reduction of material decomposition bias. The sparsely sampled spectral data are mismatched in resolution compared to the ultra-high-resolution conventional images. To address this, we employ a deep learning fusion model that integrates low-resolution spectral iodine maps with highresolution non-spectral images. This work highlights the potential for our hybrid CT sparse spectral protocol, optimized for pediatric imaging, for delivery of accurate quantitative, highresolution results which are not possible to obtain with existing technologies in their current configurations.

  PublisherDOI

• Development of a 3D-Printed Chest Respiratory Motion Phantom with Multi-Axis Compression

  2025-11-01

  articleSenior author

  Respiratory motion compensation technologies require evaluation with respiratory motion phantoms (RMPs) prior to clinical use. However, most existing RMPs are overly simplistic in structure and motion, with movement typically limited to the superior/inferior (SI) axis. This work describes the development of a deformable chest RMP featuring realistic lung structures and a novel multi-axis deformation mechanism. Attenuation data within both lungs from a patient 4DCT were converted into 3D-printer instructions using PixelPrint, which continuously varies the density of printed material to mimic patient attenuation profiles. Meanwhile the mediastinum was set to a uniform density of 50 %. Using a flexible material, thermoplastic polyurethane (TPU), the deformable lung and mediastinum phantom was 3D-printed at 0.5 x scale. The chest wall was printed using polylactic acid, with a 15 mm coronal strip of TPU to allow the chest wall to compress. To model multi-axis deformation, SI compression from the diaphragm was facilitated with a lead screw, while anterior/posterior compression was achieved with a lacing mechanism encircling the chest. The phantom was scanned at different compression levels to match patient diaphragm and chest wall motion. Deformable image registration was then performed on both patient and phantom images to obtain deformation vectors. The resulting RMP exhibited realistic lung structures and motion profiles. The SSIM between phantom and patient lungs measured 0.95 and 0.96 at the 0 and 50 % respiratory phases respectively. Mean motion vector errors measured throughout the lungs between the 50 and 0 % phases were <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\leq 2.2 \pm 1.5 ~\text{mm}$</tex> in each orthogonal direction. Realistic RMPs such as these can enable more clinically relevant assessment during development of motion compensation technologies, facilitating more effective clinical translation.

  PublisherDOI

• Abstract No. 194 In Vivo Evaluation of Spectral CT Thermometry for Non-invasive, Real-Time Temperature Monitoring of Thermal Ablation

  Journal of Vascular and Interventional Radiology · 2025-02-19

  articleOpen access
  PublisherOA PDFDOI

• Impact of vessel size, dose levels, and body habitus on iodine quantification in cardiovascular photon-counting computed tomography

  British Journal of Radiology · 2025-07-30 · 1 citations

  articleOpen accessSenior author

  OBJECTIVES: This study evaluates the performance of a clinical dual-source photon-counting computed tomography (PCCT) system in quantifying iodine within calcified vessels, using 3D-printed phantoms with vascular-like structures lined with calcium. METHODS: Parameters assessed include lumen diameters (4, 6, 8, 10, and 12 mm), phantom sizes (S: 20 × 20 cm, M: 25 × 25 cm, L: 30 × 40 cm), and iodine concentrations (2, 5, and 10 mg/mL). Scans were performed with a cardiac high-pitch acquisition protocol at radiation dose levels of 5 and 10 mGy to systematically evaluate iodine quantification accuracy and spectral imaging performance. RESULTS: The results indicate that for lumen diameters ≥6 mm, iodine quantification remains stable across all dose levels and smaller phantom sizes, where error remained consistently below 0.9 mg/mL. Furthermore, iodine quantification revealed a significant dependence on phantom size while selected radiation dose levels were insignificant. Virtual monoenergetic imaging at 70 keV showed stable performance for larger lumens (≥6 mm) with variations of 20.3 ± 13.2 HU across all conditions, while smaller lumens remained stable in medium to small phantoms. CONCLUSIONS: These findings highlight the influence of lumen diameter, patient size, and radiation dose in optimizing PCCT protocols for spectral imaging. Results indicate that PCCT maintains stable and precise imaging performance across diverse patient anatomies, with robust differentiation of iodine and calcium in adjacent regions. ADVANCES IN KNOWLEDGE: This study demonstrates PCCT's potential to enhance spectral imaging in vascular applications, characterizing iodine quantification at relevant lesion sizes for vascular imaging.

  PublisherOA PDFDOI

• Pixelprint ^PET - 3D-Printed Anthropomorphic Phantoms for Quantitative PET/CT Imaging

  2025-11-01 · 1 citations

  articleSenior author

  Advancing developments in PET / CT toward clinical implementation can be expedited through early, realistic system characterization, particularly using anatomically realistic phantoms. Accordingly, the development of phantoms that accurately simulate both radiotracer distribution and X-ray attenuation is critical for the comprehensive evaluation and optimization of PET/CT technologies. We present PixelPrint <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">${ }^{\text{PET}}$</tex>, 3D-printed phantoms for PET/CT imaging that accurately reproduce both radiotracer distribution and X-ray attenuation, with high anatomical fidelity. Calibration phantoms with various infill ratios were filled with radioactive solution and scanned using the PET/CT system at UPenn, comprising a clinical dual-layer CT scanner (IQon Spectral CT, Philips Healthcare) and the PennPET Explorer. The relationship between normalized SUV and infill ratios was assessed using linear regression. Based on the simulated whole-brain image of the BigBrain-MR dataset, a contrast ratio map of tissue FDG uptake was generated, then converted into an infill ratio map. A three-dimensional region including white matter, subthalamic and red nuclei, and cortical gray matter was selected and printed. The reconstructed images closely resemble the original input model, accurately reflecting variations in contrast and radiotracer uptake, along with realistic anatomical texture. A board-certified radiologist has reviewed and confirmed the anatomical accuracy of the brain phantom image. This work demonstrates the feasibility of anatomically accurate, multimodal phantoms that enable realistic evaluation and optimization of PET/CT systems, bridging the gap between simulation and clinical application.

  PublisherDOI

• Quantitative characterization of speckle-based x-ray imaging setup for sub-resolution microstructure analysis using standardized samples

  2025-04-08

  articleSenior author

  Chronic respiratory diseases are the third leading cause of death worldwide, affecting over 454.6 million people. Current diagnostic methods, such as imaging, often fail to detect early lung changes effectively. However, when x-rays are considered as electromagnetic waves, additional contrast mechanisms like diffraction, phase-shift, and small-angle scattering become accessible. X-ray scattering in healthy lung alveoli produce a strong dark-field signal, which decreases when alveolar integrity is compromised. This suggests dark-field imaging could be valuable for evaluating alveoli in vivo, although it remains unavailable for clinical and preclinical applications as many wave-optical imaging methods require complicated, shock-sensitive, and expensive hardware. Speckle-based dark-field imaging addresses this issue, generating dark-field contrast and transmission images using only a relatively simple diffuser material. We present a small animal lung imaging setup with a liquid-metal jet x-ray source operating at 50kV and a photon-counting detector. A sandpaper diffuser was used to create interference patterns, and polystyrene spheres (3 to100μm) were used to characterize the system. In vivo-like conditions were simulated with a single-shot technique and short acquisition times. We found that the contrast-to- noise ratio for dark-field images was significantly higher than transmission images. The results demonstrate that speckle-based dark-field imaging can detect structural changes in lung alveoli, showing a clear correlation between particle size and signal. This study underscores the potential of wave-optical methods for both preclinical and longitudinal assessment of pulmonary structures.

  PublisherDOI

• 3D PRINTING OF STACKABLE VARIABLE DENSITY RANGE MODULATORS FOR FLASH PROTON THERAPY

  International Journal of Particle Therapy · 2025-11-25

  articleOpen access
  PublisherDOI

Frequent coauthors

• Ernst J. Rummeny

  289 shared

• Franz Pfeiffer

  Institute for Advanced Study

  257 shared

• Felix K. Kopp

  Technical University of Munich

  139 shared

• Kai Mei

  University of Pennsylvania

  138 shared

• Alexander A. Fingerle

  Klinikum rechts der Isar

  138 shared

• Thomas Baum

  Technical University of Munich

  132 shared

• Jan S. Kirschke

  Technical University of Munich

  97 shared

• Thomas Koehler

  Philips (Germany)

  97 shared

Labs

• Laboratory for Advanced Computed Tomography ImagingPI