Deep Learning Reconstruction for Ultra-High-Resolution Photon-Counting CT of the Lung: Image Quality and Texture Characterization
- 1University of Pennsylvania
Abstract
BackgroundDeep learning reconstruction (DLR), in contrast to hybrid iterative reconstruction (IR), offers an alternative approach to mitigate noise and improve contrast in ultra-high-resolution (UHR) photon-counting CT (PCCT).
PurposeTo characterize image noise, lesion contrast, spatial resolution, and attenuation distribution for DLR using a patient-derived 3D-printed phantom in UHR PCCT, and to apply the same analytical framework to patient examinations acquired on the same system.
MethodsA patient CT scan-derived lung phantom containing three digitally inserted lesions of differing morphology was fabricated using PixelPrint 3D printing and imaged on a CZT-based PCCT prototype at four dose levels (CTDIvol 6, 4, 2, and 1 mGy). Images were reconstructed with hybrid IR at normal resolution (512 x 512) and UHR (1024 x 1024), and with DLR at UHR. Image noise, lesion contrast, and contrast-to-noise ratio (CNR) were measured within prespecified regions of interest replicated across reconstructions by coordinate transfer. Spatial resolution was assessed by the modulation transfer function (MTF) from a high-contrast edge at eight locations. Attenuation distribution and local texture were characterized using histograms and gray-level co-occurrence matrices (GLCMs) within a structured parenchymal region. Two patient examinations with part solid lung nodules acquired on the same prototype were analyzed as illustrative examples, with solid and ground-glass components segmented separately and their attenuation distributions compared.
ResultsIn the phantom, DLR reduced image noise relative to UHR hybrid IR (7.8% at 6 mGy and 15.2% at 1 mGy). Lesion contrast was constant across reconstructions and dose levels, so CNR differences reflected differences in noise alone. MTF curves for DLR lay above those for hybrid IR across the evaluated spatial frequency range at both 6 and 1 mGy, averaged +0.4 lp/mm in MTF50. In a structured parenchymal region, the attenuation histogram mode became progressively narrower from hybrid IR at 1 mGy to DLR at 6 mGy (averaged -23%), while mode position and mean regional attenuation were preserved, consistent with reduced partial-volume mixing rather than altered attenuation. In the patient example, the attenuation distributions of solid and ground-glass components were more widely separated with UHR DLR (2.0) than with either hybrid IR(1.6).
ConclusionsIn UHR photon-counting CT of the lung, DLR reduced image noise while preserving lesion contrast and regional attenuation, increased measured MTF at a high-contrast edge, and produced attenuation distributions consistent with reduced partial-volume mixing at material boundaries. The same effect was observed in the separation of solid and ground-glass components in two patient examples.
Related
- Impact of super resolution deep learning reconstruction with a 1024 matrix on CAD-RADS classification of intermediate coronary stenosis: A bicentric retrospective study
- Post-implantation 3-T MRI in patients with an epicranial stimulation device: image quality and deep learning-based electrode reconstruction
- Deep Learning Reconstruction versus Hybrid Iterative Reconstruction for Spectral CT: A Multi-Institutional Image Quality Assessment
- Machine learning-based electrofacies characterization of the Asmari reservoir in the Ahvaz oil field (SW Iran) using multi-resolution graph-based clustering (MRGC) and hydraulic flow unit (HFU) analysis
- Acquisition Speed versus Spatial Resolution in Ultra-High-Resolution Photon-Counting CT: Phantom Study to Guide Protocol Development
BibTeX
@misc{mei2026deep,
title = {{Deep Learning Reconstruction for Ultra-High-Resolution Photon-Counting CT of the Lung: Image Quality and Texture Characterization}},
author = {K. Mei and L. Roshkovan and L. P. Liu and S. S. Halliburton and S. Sharma and S. Ross and T. Stroud and N. Akino and Z. Yu and R. Thompson and A. H. Dhanaliwala and H. I. Litt and P. B. Noel},
howpublished = {medRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.09.04.26361840},
url = {https://www.medrxiv.org/content/10.64898/2026.09.04.26361840v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Deep Learning Reconstruction for Ultra-High-Resolution Photon-Counting CT of the Lung: Image Quality and Texture Characterization
AU - K. Mei
AU - L. Roshkovan
AU - L. P. Liu
AU - S. S. Halliburton
AU - S. Sharma
AU - S. Ross
AU - T. Stroud
AU - N. Akino
AU - Z. Yu
AU - R. Thompson
AU - A. H. Dhanaliwala
AU - H. I. Litt
AU - P. B. Noel
PY - 2026
DA - 2026/09/04
PB - medRxiv
DO - 10.64898/2026.09.04.26361840
UR - https://www.medrxiv.org/content/10.64898/2026.09.04.26361840v1
AB - BackgroundDeep learning reconstruction (DLR), in contrast to hybrid iterative reconstruction (IR), offers an alternative approach to mitigate noise and improve contrast in ultra-high-resolution (UHR) photon-counting CT (PCCT). PurposeTo characterize image noise, lesion contrast, spatial resolution, and attenuation distribution for DLR using a patient-derived 3D-printed phantom in UHR PCCT, and to apply the same analytical framework to patient examinations acquired on the same system. MethodsA patient CT scan-derived lung phantom containing three digitally inserted lesions of differing morphology was fabricated using PixelPrint 3D printing and imaged on a CZT-based PCCT prototype at four dose levels (CTDIvol 6, 4, 2, and 1 mGy). Images were reconstructed with hybrid IR at normal resolution (512 x 512) and UHR (1024 x 1024), and with DLR at UHR. Image noise, lesion contrast, and contrast-to-noise ratio (CNR) were measured within prespecified regions of interest replicated across reconstructions by coordinate transfer. Spatial resolution was assessed by the modulation transfer function (MTF) from a high-contrast edge at eight locations. Attenuation distribution and local texture were characterized using histograms and gray-level co-occurrence matrices (GLCMs) within a structured parenchymal region. Two patient examinations with part solid lung nodules acquired on the same prototype were analyzed as illustrative examples, with solid and ground-glass components segmented separately and their attenuation distributions compared. ResultsIn the phantom, DLR reduced image noise relative to UHR hybrid IR (7.8% at 6 mGy and 15.2% at 1 mGy). Lesion contrast was constant across reconstructions and dose levels, so CNR differences reflected differences in noise alone. MTF curves for DLR lay above those for hybrid IR across the evaluated spatial frequency range at both 6 and 1 mGy, averaged +0.4 lp/mm in MTF50. In a structured parenchymal region, the attenuation histogram mode became progressively narrower from hybrid IR at 1 mGy to DLR at 6 mGy (averaged -23%), while mode position and mean regional attenuation were preserved, consistent with reduced partial-volume mixing rather than altered attenuation. In the patient example, the attenuation distributions of solid and ground-glass components were more widely separated with UHR DLR (2.0) than with either hybrid IR(1.6). ConclusionsIn UHR photon-counting CT of the lung, DLR reduced image noise while preserving lesion contrast and regional attenuation, increased measured MTF at a high-contrast edge, and produced attenuation distributions consistent with reduced partial-volume mixing at material boundaries. The same effect was observed in the separation of solid and ground-glass components in two patient examples.
ER -