Academic Bibliography

 Search 200 years of publications by Ghent University researchers.
Advanced search
2 files | 8.39 MB
Add to list
  1. Search results
  2. Cross-modality deep learning denoisin...

Cross-modality deep learning denoising for low-dose μSPECT : transfer of PET-trained U-net and diffusion models

Elise Taragola (UGent) , Sara Neyt (UGent) , Boxiao Yu, Maya Abi Akl (UGent) , Boris Vervenne (UGent) , Kuang Gong, Stefaan Vandenberghe (UGent) , Christian Vanhove (UGent) and Florence Marie Muller (UGent)
(2026) CHEMICAL & BIOMEDICAL IMAGING.
Author
Organization
Project
Abstract
Reducing radiation dose in microsingle photon emission computed tomography (μSPECT) is essential to limit adverse biological effects in small animal studies, especially for longitudinal experiments. However, dose reduction increases noise and degrades image quality. While deep learning-based denoising (DL-DN) has shown promise for low-dose imaging, most prior work has focused on positron emission tomography (PET). DL-DN for μSPECT remains largely unexplored. In this study, we evaluate postreconstruction DL-DN for low-count μSPECT using two 2D architectures: a U-Net and a denoising diffusion probabilistic model (DDPM). To mitigate the limited availability of preclinical training data, we investigated transfer-learning from PET, a related molecular imaging modality, using models pretrained on both μPET (mouse) and clinical PET (human) data. These models were evaluated in a zero-shot setting and further adapted to the μSPECT domain via transfer-learning. Their performance was compared to models trained from scratch using varying amounts of μSPECT data (4–32 volumes). The data set consisted of in vivo mouse scans acquired with two tracers at 10%, 25%, and 50% of standard counts. Both architectures achieved meaningful noise reduction and improved quantitative agreement with standard-count references. Zero-shot PET-pretrained models demonstrated cross-modality generalization, reducing root mean squared error (RMSE) by more than 25% at the lowest count level, although residual artifacts remained due to differences between the PET source and μSPECT target domains. Transfer-learning consistently improved performance, yielding additional RMSE reductions exceeding 5% relative to zero-shot models. Its advantage over scratch-training decreased with increasing data set size, with maximal RMSE gains dropping from over 10% for small training sets to below 3% for larger data sets. Overall, the U-Net provided more robust performance with lower computational cost. These results indicate that DL-DN can support dose reduction in μSPECT, and that transfer-learning from PET offers a practical approach to address the small data set sizes common for preclinical settings.
Keywords
Deep learning, Denoising diffusion probabilistic models, Dose reduction, Image denoising, Micro-SPECT, Transfer learning

Downloads

  • Download
    cross-modality-deep-learning-denoising-for-low-dose-%CE%BCspect-transfer-of-pet-trained-u-net-and-diffusion-models.pdf
    • full text (Published version)
    • |
    • open access
    • |
    • PDF
    • |
    • 6.13 MB
  • Download
    im5c00270 si 001.pdf
    • supplementary material
    • |
    • open access
    • |
    • PDF
    • |
    • 2.26 MB

Citation

Please use this url to cite or link to this publication:

MLA
Taragola, Elise, et al. “Cross-Modality Deep Learning Denoising for Low-Dose ΜSPECT : Transfer of PET-Trained U-Net and Diffusion Models.” CHEMICAL & BIOMEDICAL IMAGING, edited by Deju Ye, 2026, doi:10.1021/cbmi.5c00270.
APA
Taragola, E., Neyt, S., Yu, B., Abi Akl, M., Vervenne, B., Gong, K., … Muller, F. M. (2026). Cross-modality deep learning denoising for low-dose μSPECT : transfer of PET-trained U-net and diffusion models. CHEMICAL & BIOMEDICAL IMAGING. https://doi.org/10.1021/cbmi.5c00270
Chicago author-date
Taragola, Elise, Sara Neyt, Boxiao Yu, Maya Abi Akl, Boris Vervenne, Kuang Gong, Stefaan Vandenberghe, Christian Vanhove, and Florence Marie Muller. 2026. “Cross-Modality Deep Learning Denoising for Low-Dose ΜSPECT : Transfer of PET-Trained U-Net and Diffusion Models.” Edited by Deju Ye. CHEMICAL & BIOMEDICAL IMAGING. https://doi.org/10.1021/cbmi.5c00270.
Chicago author-date (all authors)
Taragola, Elise, Sara Neyt, Boxiao Yu, Maya Abi Akl, Boris Vervenne, Kuang Gong, Stefaan Vandenberghe, Christian Vanhove, and Florence Marie Muller. 2026. “Cross-Modality Deep Learning Denoising for Low-Dose ΜSPECT : Transfer of PET-Trained U-Net and Diffusion Models.” Ed by. Deju Ye. CHEMICAL & BIOMEDICAL IMAGING. doi:10.1021/cbmi.5c00270.
Vancouver
1.
Taragola E, Neyt S, Yu B, Abi Akl M, Vervenne B, Gong K, et al. Cross-modality deep learning denoising for low-dose μSPECT : transfer of PET-trained U-net and diffusion models. Ye D, editor. CHEMICAL & BIOMEDICAL IMAGING. 2026;
IEEE
[1]
E. Taragola et al., “Cross-modality deep learning denoising for low-dose μSPECT : transfer of PET-trained U-net and diffusion models,” CHEMICAL & BIOMEDICAL IMAGING, 2026.
@article{01KMN42BSWTP240A22XGVXPZZG,
  abstract     = {{Reducing radiation dose in microsingle photon emission computed tomography (μSPECT) is essential to limit adverse biological effects in small animal studies, especially for longitudinal experiments. However, dose reduction increases noise and degrades image quality. While deep learning-based denoising (DL-DN) has shown promise for low-dose imaging, most prior work has focused on positron emission tomography (PET). DL-DN for μSPECT remains largely unexplored. In this study, we evaluate postreconstruction DL-DN for low-count μSPECT using two 2D architectures: a U-Net and a denoising diffusion probabilistic model (DDPM). To mitigate the limited availability of preclinical training data, we investigated transfer-learning from PET, a related molecular imaging modality, using models pretrained on both μPET (mouse) and clinical PET (human) data. These models were evaluated in a zero-shot setting and further adapted to the μSPECT domain via transfer-learning. Their performance was compared to models trained from scratch using varying amounts of μSPECT data (4–32 volumes). The data set consisted of in vivo mouse scans acquired with two tracers at 10%, 25%, and 50% of standard counts. Both architectures achieved meaningful noise reduction and improved quantitative agreement with standard-count references. Zero-shot PET-pretrained models demonstrated cross-modality generalization, reducing root mean squared error (RMSE) by more than 25% at the lowest count level, although residual artifacts remained due to differences between the PET source and μSPECT target domains. Transfer-learning consistently improved performance, yielding additional RMSE reductions exceeding 5% relative to zero-shot models. Its advantage over scratch-training decreased with increasing data set size, with maximal RMSE gains dropping from over 10% for small training sets to below 3% for larger data sets. Overall, the U-Net provided more robust performance with lower computational cost. These results indicate that DL-DN can support dose reduction in μSPECT, and that transfer-learning from PET offers a practical approach to address the small data set sizes common for preclinical settings.}},
  author       = {{Taragola, Elise and Neyt, Sara and Yu, Boxiao and Abi Akl, Maya and Vervenne, Boris and Gong, Kuang and Vandenberghe, Stefaan and Vanhove, Christian and Muller, Florence Marie}},
  editor       = {{Ye, Deju}},
  issn         = {{2832-3637}},
  journal      = {{CHEMICAL & BIOMEDICAL IMAGING}},
  keywords     = {{Deep learning,Denoising diffusion probabilistic models,Dose reduction,Image denoising,Micro-SPECT,Transfer learning}},
  language     = {{eng}},
  pages        = {{13}},
  title        = {{Cross-modality deep learning denoising for low-dose μSPECT : transfer of PET-trained U-net and diffusion models}},
  url          = {{http://doi.org/10.1021/cbmi.5c00270}},
  year         = {{2026}},
}
Altmetric
View in Altmetric