Impact of a Deep Learning Reconstruction Algorithm Combined with an Ultra-high Resolution Detector on the Quality of Temporomandibular Joint Computed Tomography Images
CT Lilun yu yingyong yanjiu · 2026-07 · 35 (4) · 689-694
DOI: 10.15953/j.ctta.2025.482 ↗Kurz gefasst
Fachartikel in der Zeitschrift CT Lilun yu yingyong yanjiu, 2026-07. Verfasst von Qing ZHENG und 5 weiteren Autor:innen.
Abstract (Originalsprache)
Objective
This study aims to explore the impact of an ultra-high resolution detector combined with a deep learning reconstruction algorithm, ClearInfinity (CI), on the image quality of temporomandibular joint (TMJ) computed tomography (CT) scans.
Methods
Seven fresh cadaveric head specimens were scanned using the NeuViz Epoch Elite CT scanner, with two ultra-high resolution collimation widths (76×0.156 mm and 128×0.3125 mm). For each collimation width, filtered back projection (FBP), adaptive iterative reconstruction ClearView (CV) 60%, and deep learning reconstruction algorithm CI60% were applied, resulting in six sets of images. The image quality was assessed by comparing the contrast-to-noise ratio (CNR), signal-to-noise ratio (SNR), and subjective scores using a double-blind method, followed by a statistical analysis.
Results
In the objective evaluation, the CNR of the condyle and joint for the 0.156 mm collimation width images were significantly higher than those for the 0.3125 mm collimation width images. The CNR of the condyle and SNRs of the condyle joint in the CI60% images were significantly higher than those obtained using other algorithms. In the subjective evaluation, the scores for the condylar cortical bone and joint fossa in the 0.156 mm collimation width images were significantly higher than those in the 0.3125 mm collimation width images. The subjective scores for condylar cortical bone, trabecular bone, joint eminence, and joint fossa in the CI60% images were significantly higher than those obtained using other algorithms, and the inter-rater consistency was good (Kappa\begin{document}$\geq $\end{document}0.75).
Conclusion
The 0.156 mm ultra-high resolution detector combined with the deep learning algorithm CI could significantly improve the resolution and contrast of TMJ CT images, reduce noise and artifacts, and show promising clinical application prospects.
Schlagwörter
Zitieren
APA
ZHENG, Q., MA, Z., SU, Y., SONG, L., ZHANG, Y., & LIU, D. (2026). Impact of a Deep Learning Reconstruction Algorithm Combined with an Ultra-high Resolution Detector on the Quality of Temporomandibular Joint Computed Tomography Images. CT Lilun yu yingyong yanjiu, 35(4), 689-694. https://doi.org/10.15953/j.ctta.2025.482
Vancouver
ZHENG Q, MA Z, SU Y, SONG L, ZHANG Y, LIU D. Impact of a Deep Learning Reconstruction Algorithm Combined with an Ultra-high Resolution Detector on the Quality of Temporomandibular Joint Computed Tomography Images. CT Lilun yu yingyong yanjiu. 2026;35(4):689-694. doi:10.15953/j.ctta.2025.482
BibTeX
@article{ZHENG2026,
author = {ZHENG, Qing and MA, Zixuan and SU, Yaping and SONG, Liyuan and ZHANG, Yongxian and LIU, Dandan},
title = {Impact of a Deep Learning Reconstruction Algorithm Combined with an Ultra-high Resolution Detector on the Quality of Temporomandibular Joint Computed Tomography Images},
journal = {CT Lilun yu yingyong yanjiu},
year = {2026},
volume = {35},
number = {4},
pages = {689--694},
doi = {10.15953/j.ctta.2025.482}
}
EUVIMED