Statistically Conditioned MRI Denoising via Film-Modulated Residual Attention U-Net
Abstract
The quality of MRI images is often limited by spatially inhomogeneous noise, which negatively affects the accuracy of clinical interpretation and automatic analysis. Traditional deep learning methods often implicitly account for noise, leading to excessive smoothing and the loss of fine anatomical structures. In this paper, we propose an Enhanced Denoising U-Net architecture that employs a Feature-wise Linear Modulation (FiLM) mechanism to dynamically adapt to the noise profile of each slice. The model combines a vector of 8 statistical descriptors (including intensity, texture, and frequency characteristics), enabling dynamic control of the network’s internal representations based on specific scanning conditions. To improve physical correctness, training was performed on data with synthetically generated k-space noise. The architecture is enhanced with residual blocks, attention mechanisms, and a multiscale processing module. On synthetic data, the average Peak Signal-to-Noise Ratio (PSNR) improvement was ≈ 20.7 dB, and with an average Structural Similarity Index (SSIM) improvement of approximately 0.73, indicating a deep restoration of structural information. In clinical images, an increase in SNR and stabilization of the coefficient of variation (CV) were observed, confirming the method's physical correctness. Clinical validation on complex contoured structures (hippocampus, brainstem, optic chiasm) showed an increase in the Dice coefficient (DSC) by 0.07–0.12 and a decrease in the HD95 error by 30–50%. The proposed method enables a transition from universal denoising strategies to adaptive reconstruction, ensuring high accuracy of preserving anatomical boundaries. This makes it a promising tool for MRI processing in neuroimaging tasks and variable therapy planning.
Downloads
References
F. Lugauer, and J. Wetzl, “Magnetic Resonance Imaging,” in: Medical Imaging Systems: An Introductory Guide, edited by A. Maier, S. Steidl, V. Christlein, and J. Hornegger, (Springer, Cham (CH), 2018). https://doi.org/10.1007/978-3-319-96520-8_6
C.M. Sonderer, and N. Chen, “Improving the Accuracy, Quality, and Signal-To-Noise Ratio of MRI Parametric Mapping Using Rician Bias Correction and Parametric-Contrast-Matched Principal Component Analysis (PCM-PCA),” Yale J Biol Med 91(3), 207–214 (2018). PMCID: PMC6153621
W.A. Edelstein, G.H. Glover, C.J. Hardy, and R.W. Redington, “The intrinsic signal-to-noise ratio in NMR imaging,” Magn Reson Med 3(4), 604–618 (1986). https://doi.org/10.1002/mrm.1910030413
H. Gudbjartsson, and S. Patz, “The Rician distribution of noisy MRI data,” Magn. Reson. Med. 34(6), 910–914 (1995). https://doi.org/10.1002/mrm.1910340618
V. Prabhakaran, V.A. Nair, B.P. Austin, C. La, T.A. Gallagher, Y. Wu, D.G. McLaren, G. Xu, P. Turski, and H. Rowley, “Current Status and Future Perspectives of Magnetic Resonance High-Field Imaging: A Summary,” Neuroimaging Clin. N. Am. 22(2), 373–xii (2012). https://doi.org/10.1016/j.nic.2012.02.012
K. Dabov, A. Foi, V. Katkovnik, and K. Egiazarian, “Image Denoising by Sparse 3-D Transform-Domain Collaborative Filtering,” IEEE Transactions on Image Processing, 16(8), 2080–2095 (2007). https://doi.org/10.1109/TIP.2007.901238
M. Elad, B. Kawar, and G. Vaksman, “Image Denoising: The Deep Learning Revolution and Beyond—A Survey Paper,” SIAM Journal on Imaging Sciences,” 16(13), 1594-1654 (2023). https://doi.org/10.1137/23M1545859
S.W. Zamir, A. Arora, S. Khan, M. Hayat, F.S. Khan, M.-H. Yang, and L. Shao, “Learning Enriched Features for Real Image Restoration and Enhancement,” in: Computer Vision – ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XXV, (Springer-Verlag, Berlin, Heidelberg, 2020), pp. 492–511. https://doi.org/10.1007/978-3-030-58595-2_30
J. Lehtinen, J. Munkberg, J. Hasselgren, S. Laine, T. Karras, M. Aittala, and T. Aila, “Noise2Noise: Learning image restoration without clean data,” in: 35th International Conference on Machine Learning, ICML 2018, (International Machine Learning Society, 2018), pp. 4620–4631. https://doi.org/10.48550/arXiv.1803.04189
E. Perez, F. Strub, H. de Vries, V. Dumoulin, and A. Courville, “FiLM: visual reasoning with a general conditioning layer,” in: Proceedings of the Thirty-Second AAAI Conference on Artificial Intelligence and Thirtieth Innovative Applications of Artificial Intelligence Conference and Eighth AAAI Symposium on Educational Advances in Artificial Intelligence, (AAAI Press, New Orleans, Louisiana, USA, 2018), pp. 3942–3951. https://doi.org/10.48550/arXiv.1709.07871
O. Oktay, J. Schlemper, L.L. Folgoc, M. Lee, M. Heinrich, K. Misawa, K. Mori, S. McDonagh, N.Y. Hammerla, B. Kainz, B. Glocker, and D. Rueckert, “Attention U-Net: Learning Where to Look for the Pancreas,” (2022). https://doi.org/10.48550/arXiv.1804.03999
X. Chen, X. Wang, K. Zhang, K.-M. Fung, T.C. Thai, K. Moore, R.S. Mannel, H. Liu, B. Zheng, and Y. Qiu, “Recent advances and clinical applications of deep learning in medical image analysis,” Medical Image Analysis 79, 102444 (2022). https://doi.org/10.1016/j.media.2022.102444
R. Ayde, M. Vornehm, Y. Zhao, F. Knoll, E.X. Wu, and M. Sarracanie, “MRI at low field: A review of software solutions for improving SNR,” NMR Biomed, 38(1), e5268 (2025). https://doi.org/10.1002/nbm.5268
M. Safari, S. Wang, Z. Eidex, R. Qiu, C.-W. Chang, D.S. Yu, and X. Yang, “A Physics-Informed Deep Learning Model for MRI Brain Motion Correction,” ArXiv, arXiv:2502.09296v1 (2025). https://doi.org/10.1002/mp.70197
L. Han, S. Xiao, M. Li, J. Liu, and X. Zhou, “Noise-Controllable Complex-Valued Diffusion Model for k-Space Data of Hyperpolarized $$^{129}$$Xe Lung MRI Generation,” in: Medical Image Computing and Computer Assisted Intervention – MICCAI 2025, edited by J.C. Gee, D.C. Alexander, J. Hong, J.E. Iglesias, C.H. Sudre, A. Venkataraman, P. Golland, J.H. Kim, and J. Park, (Springer Nature Switzerland, Cham, 2026), pp. 360–369. https://doi.org/10.1007/978-3-032-05127-1_35
C. Ciulla, “Two-dimensional image noise removal and reconstruction using discrete Fourier transform, k-space filtering and Z-space filtering,” Progress in Engineering Science, 2(1), 100056 (2025). https://doi.org/10.1016/j.pes.2025.100056
T. Song, F. Nie, Y. Guo, F. Xu, and S. Zhang, “FilterDiff: Noise-Free Frequency-Domain Diffusion Models for Accelerated MRI Reconstruction,” in: Medical Image Computing and Computer Assisted Intervention – MICCAI 2025, edited by J.C. Gee, D.C. Alexander, J. Hong, J.E. Iglesias, C.H. Sudre, A. Venkataraman, P. Golland, J.H. Kim, and J. Park, (Springer Nature Switzerland, Cham, 2026), pp. 205–215. https://doi.org/10.1007/978-3-032-05325-1_20
W. Zhang, Y. Guo, and Q. Jin, “Radiomics and Its Feature Selection: A Review,” Symmetry, 15(10), 1834 (2023). https://doi.org/10.3390/sym15101834
M. Shakeri, A. Mostaar, A.Z. Sadeghi, S.M. Hosseini, A.Y. Joybari, and H. Ghadiri, “A Comprehensive Evaluation of Radiomic Features in Normal Brain Magnetic Resonance Imaging: Investigating Robustness and Region Variations,” J. Med. Phys. 49(4), 608–622 (2024). https://doi.org/10.4103/jmp.jmp_149_24
M. Dohmen, M.A. Klemens, I.M. Baltruschat, T. Truong, and M. Lenga, “Similarity and quality metrics for MR image-to-image translation,” Sci. Rep. 15(1), 3853 (2025). https://doi.org/10.1038/s41598-025-87358-0
“The Image Biomarker Standardization Initiative: Standardized Quantitative Radiomics for High-Throughput Image-based Phenotyping,” Radiology,” 295(2), (2020). https://doi.org/10.1148/radiol.2020191145
D. Sliusarenko, and A. Netreba, “Phantom-Guided Adaptive Denoising of Medical Images Using Enhanced U-Net Architecture,” Radioelectronics and Communications Systems, (2025). https://doi.org/10.3103/S0735272725020049
M. Mathieu, C. Couprie, and Y. LeCun, “Deep multi-scale video prediction beyond mean square error,” CoRR, (2015). https://doi.org/10.48550/arXiv.1511.05440
P. Isola, J.-Y. Zhu, T. Zhou, and A.A. Efros, “Image-to-Image Translation with Conditional Adversarial Networks,” in: 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), (2017), pp. 5967–5976. https://doi.org/10.1109/CVPR.2017.632
D. Sliusarenko, A. Netreba, and S. Radchenko, “Refining Clarity Medical Image via U-Net with Tailored Architectural Modifications and Transfer Learning,” in: Proc. IEEE Int. Conf. Electron. Nanotechnol., ELNANO., (Institute of Electrical and Electronics Engineers, 2024), pp. 456–460. https://doi.org/10.1109/ELNANO63394.2024.10756858
H. Liu, Z. Li, S. Lin, and L. Cheng, “A Residual UNet Denoising Network Based on Multi-Scale Feature Extraction and Attention-Guided Filter,” Sensors (Basel), 23(16), 7044 (2023). https://doi.org/10.3390/s23167044
H. Zhang, Q. Lian, J. Zhao, Y. Wang, Y. Yang, and S. Feng, “RatUNet: residual U-Net based on attention mechanism for image denoising,” PeerJ Comput. Sci. 8, e970 (2022). https://doi.org/10.7717/peerj-cs.970
Z. Zhuang, M. Liu, A. Cutkosky, and F. Orabona, “Understanding AdamW through Proximal Methods and Scale-Freeness,” Trans. Mach. Learn. Res. (2022). https://doi.org/10.48550/arXiv.2202.00089
S. Brody, U. Alon, and E. Yahav, “On the Expressivity Role of LayerNorm in Transformers’ Attention,” (arXiv, 2023). https://doi.org/10.48550/arXiv.2305.02582
T.-T. Han, H. Nguyen Van, and P. Nguyen Huu, “Denoising Method for MRI Images Using Modified BM3D Filter with Complex Network and Artificial Neural Networks,” Journal of Electrical and Computer Engineering, 2024(1), 2606485 (2024). https://doi.org/10.1155/2024/2606485
Z. Wang, A.C. Bovik, H.R. Sheikh, and E.P. Simoncelli, “Image quality assessment: from error visibility to structural similarity,” IEEE Transactions on Image Processing, 13(4), 600–612 (2004). https://doi.org/10.1109/TIP.2003.819861
M.J. Kobra, M.O. Rahman, and A.M. Nakib, “A Novel Hybrid Framework for Noise Estimation in High-Texture Images using Markov, MLE, and CNN Approaches,” Scientific Journal of Engineering Research, 1(2), 54–63 (2025). https://doi.org/10.64539/sjer.v1i2.2025.25
H.M.S.S. Herath, H.M.K.K.M.B. Herath, N. Madusanka, and B.-I. Lee, “A Systematic Review of Medical Image Quality Assessment,” Journal of Imaging, 11(4), 100 (2025). https://doi.org/10.3390/jimaging11040100
R. Rodrigues, L. Lévêque, J. Gutiérrez, H. Jebbari, M. Outtas, L. Zhang, A. Chetouani, S. Al-Juboori, M.G. Martini, and A.M.G. Pinheiro, “Objective quality assessment of medical images and videos: review and challenges,” Multimed Tools Appl 84(25), 29915–29948 (2025). https://doi.org/10.1007/s11042-024-20292-x
K.M. Kempski, M.T. Graham, M.R. Gubbi, T. Palmer, and M.A. Lediju Bell, “Application of the generalized contrast-to-noise ratio to assess photoacoustic image quality,” Biomed. Opt. Express, 11(7), 3684–3698 (2020). https://doi.org/10.1364/BOE.391026
O. Rainio, and R. Klén, “Modified Dice Coefficients for Evaluation of Tumor Segmentation from PET Images: A Proof-of-Concept Study,” J. Digit. Imaging. Inform. Med. (2025). https://doi.org/10.1007/s10278-025-01535-1
P. Ramachandran, T. Eswarlal, M. Lehman, and Z. Colbert, “Assessment of Optimizers and their Performance in Autosegmenting Lung Tumors,” J. Med. Phys. 48(2), 129–135 (2023). https://doi.org/10.4103/jmp.jmp_54_23
A. Mason, J. Rioux, S.E. Clarke, A. Costa, M. Schmidt, V. Keough, T. Huynh, and S. Beyea, “Comparison of Objective Image Quality Metrics to Expert Radiologists’ Scoring of Diagnostic Quality of MR Images,” IEEE Transactions on Medical Imaging, 39(4), 1064–1072 (2020). https://doi.org/10.1109/TMI.2019.2930338
K. Chupetlovska, K. Groot Lipman, Z. Bodalal, F.M. Aricò, L. Topff, M. Maas, and S. Trebeschi, “Do’s and don’ts of tumor segmentation with 3D slicer: A practical guide for radiologists, by radiologists,” European Journal of Radiology Artificial Intelligence, 5, 100053 (2026). https://doi.org/10.1016/j.ejrai.2025.100053
A. Buades, B. Coll, and J.-M. Morel, “A non-local algorithm for image denoising,” in 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR’05), (2005), pp. 60–65 vol. 2. https://doi.org/10.1109/CVPR.2005.38
K. Zhang, W. Zuo, Y. Chen, D. Meng, and L. Zhang, “Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising,” IEEE Transactions on Image Processing, 26(7), 3142–3155 (2017). https://doi.org/10.1109/TIP.2017.2662206
Q. Yang, P. Yan, Y. Zhang, H. Yu, Y. Shi, X. Mou, M.K. Kalra, Y. Zhang, L. Sun, and G. Wang, “Low Dose CT Image Denoising Using a Generative Adversarial Network with Wasserstein Distance and Perceptual Loss,” IEEE Trans Med Imaging, 37(6), 1348–1357 (2018). https://doi.org/10.1109/TMI.2018.2827462
Copyright (c) 2026 D.G. Sliusarenko, L.V. Sayan, A.V. Netreba
This work is licensed under a Creative Commons Attribution 4.0 International License.
Authors who publish with this journal agree to the following terms:
- Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution License that allows others to share the work with an acknowledgment of the work's authorship and initial publication in this journal.
- Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgment of its initial publication in this journal.
- Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work (See The Effect of Open Access).
