SVD-SN2N: Self-Inspired Noise2Noise Learning for Denoising Log-Compressed SVD-Filtered Ultrasound Imaging
- 1Physics for Medicine Paris, INSERM U1273, ESPCI Paris, PSL University, CNRS
Abstract
Singular value decomposition (SVD) is the reference clutter rejection strategy for medical ultrafast ultrasound imaging, yet SVD filtered images retain a residual noise floor that obscures microvascular signals at depth. Supervised denoising cannot address this gap because clean references do not exist, and most physics-based alternatives require radio-frequency or in-phase/quadrature data that clinical scanners do not expose. We introduce SVD SN2N, a self supervised Noise2Noise framework that operates entirely on the post SVD image and requires neither clean targets nor channel level access. Our central contribution is to show that the residual fluctuations of SVD filtered images are dominated by multiplicative speckle, which violates the zero mean additive assumption of Noise2Noise theory, and that a single log compression step converts it into an additive perturbation whose variance no longer depends on the signal, restoring the conditions the framework requires. The log compressed image is split into two statistically independent half size copies by diagonal 2 times 2 resampling, rescaled by Fourier zero padding interpolation, and used to train a U Net under a self constrained twin-prediction loss. We assess the underlying assumptions directly by measuring the residual noise distribution, its spatial correlation and the signal similarity within diagonal pixel pairs, and we validate the framework on four heterogeneous datasets spanning preclinical and clinical, contrast enhanced and contrast-free, and 2D to 3D regimes. Relative to conventional SVD, SVD SN2N raises image derived SNR by 4.2 to10.6 dB and narrows the apparent vessel full width at half maximum by 34% to 59%, providing a post SVD denoising front end compatible with the image only data available on clinical scanners.
Related
- Optimization of Deep Learning Model for Denoising Dose Profiles in Proton Therapy
- Impaired value estimation over space explains turn-induced freezing of gait in Parkinson's disease: A cortico-basal ganglia-inspired reinforcement learning model
- Adversarial robustness of deep learning-based thyroid nodule segmentation in ultrasound
- Denoising single-cell ATAC-seq data with Shannon entropy
- 2-Dimensional spectral gating for denoising bioacoustics recordings
BibTeX
@misc{zhang2026svdsn2n,
title = {{SVD-SN2N: Self-Inspired Noise2Noise Learning for Denoising Log-Compressed SVD-Filtered Ultrasound Imaging}},
author = {G. Zhang and E. Buffle and N. S. J. ZUCKER and T. Deffieux and N. Ialy-radio and M. Pernot and S. Pezet and M. Tanter},
howpublished = {bioRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.09.10.750668},
url = {https://www.biorxiv.org/content/10.64898/2026.09.10.750668v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - SVD-SN2N: Self-Inspired Noise2Noise Learning for Denoising Log-Compressed SVD-Filtered Ultrasound Imaging
AU - G. Zhang
AU - E. Buffle
AU - N. S. J. ZUCKER
AU - T. Deffieux
AU - N. Ialy-radio
AU - M. Pernot
AU - S. Pezet
AU - M. Tanter
PY - 2026
DA - 2026/09/10
PB - bioRxiv
DO - 10.64898/2026.09.10.750668
UR - https://www.biorxiv.org/content/10.64898/2026.09.10.750668v1
AB - Singular value decomposition (SVD) is the reference clutter rejection strategy for medical ultrafast ultrasound imaging, yet SVD filtered images retain a residual noise floor that obscures microvascular signals at depth. Supervised denoising cannot address this gap because clean references do not exist, and most physics-based alternatives require radio-frequency or in-phase/quadrature data that clinical scanners do not expose. We introduce SVD SN2N, a self supervised Noise2Noise framework that operates entirely on the post SVD image and requires neither clean targets nor channel level access. Our central contribution is to show that the residual fluctuations of SVD filtered images are dominated by multiplicative speckle, which violates the zero mean additive assumption of Noise2Noise theory, and that a single log compression step converts it into an additive perturbation whose variance no longer depends on the signal, restoring the conditions the framework requires. The log compressed image is split into two statistically independent half size copies by diagonal 2 times 2 resampling, rescaled by Fourier zero padding interpolation, and used to train a U Net under a self constrained twin-prediction loss. We assess the underlying assumptions directly by measuring the residual noise distribution, its spatial correlation and the signal similarity within diagonal pixel pairs, and we validate the framework on four heterogeneous datasets spanning preclinical and clinical, contrast enhanced and contrast-free, and 2D to 3D regimes. Relative to conventional SVD, SVD SN2N raises image derived SNR by 4.2 to10.6 dB and narrows the apparent vessel full width at half maximum by 34% to 59%, providing a post SVD denoising front end compatible with the image only data available on clinical scanners.
ER -