Fourier Diffusion Models: A Method to Control MTF and NPS in Score-Based Stochastic Image Generation.
Gespeichert in:
| Titel: | Fourier Diffusion Models: A Method to Control MTF and NPS in Score-Based Stochastic Image Generation. |
|---|---|
| Autoren: | Tivnan M, Teneggi J, Lee TC, Zhang R, Boedeker K, Cai L, Gang GJ, Sulam J, Stayman JW |
| Quelle: | IEEE transactions on medical imaging [IEEE Trans Med Imaging] 2025 Sep; Vol. 44 (9), pp. 3694-3704. |
| Publikationsart: | Journal Article |
| Sprache: | English |
| Info zur Zeitschrift: | Publisher: Institute of Electrical and Electronics Engineers Country of Publication: United States NLM ID: 8310780 Publication Model: Print Cited Medium: Internet ISSN: 1558-254X (Electronic) Linking ISSN: 02780062 NLM ISO Abbreviation: IEEE Trans Med Imaging Subsets: MEDLINE |
| Imprint Name(s): | Original Publication: New York, NY : Institute of Electrical and Electronics Engineers, c1982- |
| MeSH-Schlagworte: | Fourier Analysis* , Image Processing, Computer-Assisted*/methods , Tomography, X-Ray Computed*/methods, Humans ; Stochastic Processes ; Lung/diagnostic imaging ; Algorithms ; Neural Networks, Computer ; Databases, Factual |
| Abstract: | Score-based diffusion models are new and powerful tools for image generation. They are based on a forward stochastic process where an image is degraded with additive white noise and optional input scaling. A neural network can be trained to estimate the time-dependent score function, and used to run the reverse-time stochastic process to generate new samples from the training image distribution. However, one issue is that sampling the reverse process requires many passes of the neural network. In this work we present Fourier Diffusion Models which replace the scalar operations of the forward process with linear shift invariant systems and additive spatially-stationary noise. This allows for a model of continuous probability flow from true images to measurements with a specific modulation transfer function (MTF) and noise power spectrum (NPS). We also derive the reverse process for posterior sampling of high-quality images given blurry noisy measurements. We conducted a computational experiment using the Lung Image Database Consortium dataset of chest CT images and simulated CT measurements with correlated noise and system blur. Our results show that Fourier diffusion models can improve image quality for supervised diffusion posterior sampling relative to existing conditional diffusion models. |
| References: | Med Image Comput Comput Assist Interv. 2024 Oct;15010:449-458. (PMID: 40166669) IEEE Trans Med Imaging. 2021 Nov;40(11):3249-3260. (PMID: 33950837) Med Image Anal. 2022 Aug;80:102479. (PMID: 35696876) Med Phys. 2024 Mar;51(3):1847-1859. (PMID: 37646491) Comput Biol Med. 2024 Jan;168:107803. (PMID: 38064854) Med Phys. 2012 Aug;39(8):5145-56. (PMID: 22894440) Eur J Nucl Med Mol Imaging. 2024 Jan;51(2):358-368. (PMID: 37787849) IEEE Trans Pattern Anal Mach Intell. 2023 Jan;45(1):87-110. (PMID: 35180075) Small Methods. 2025 Mar;9(3):e2400672. (PMID: 39400948) Med Image Anal. 2023 Aug;88:102846. (PMID: 37295311) AMIA Annu Symp Proc. 2024 Jan 11;2023:624-633. (PMID: 38222387) IEEE Trans Neural Netw Learn Syst. 2025 Jul;36(7):11793-11813. (PMID: 39471123) IEEE Trans Pattern Anal Mach Intell. 2023 Sep;45(9):10850-10869. (PMID: 37030794) |
| Grant Information: | R01 CA249538 United States CA NCI NIH HHS |
| Entry Date(s): | Date Created: 20250321 Date Completed: 20250916 Latest Revision: 20251117 |
| Update Code: | 20251117 |
| PubMed Central ID: | PMC12619680 |
| DOI: | 10.1109/TMI.2025.3553805 |
| PMID: | 40117140 |
| Datenbank: | MEDLINE |
Full Text Finder 
Nájsť tento článok vo Web of Science | Abstract: | Score-based diffusion models are new and powerful tools for image generation. They are based on a forward stochastic process where an image is degraded with additive white noise and optional input scaling. A neural network can be trained to estimate the time-dependent score function, and used to run the reverse-time stochastic process to generate new samples from the training image distribution. However, one issue is that sampling the reverse process requires many passes of the neural network. In this work we present Fourier Diffusion Models which replace the scalar operations of the forward process with linear shift invariant systems and additive spatially-stationary noise. This allows for a model of continuous probability flow from true images to measurements with a specific modulation transfer function (MTF) and noise power spectrum (NPS). We also derive the reverse process for posterior sampling of high-quality images given blurry noisy measurements. We conducted a computational experiment using the Lung Image Database Consortium dataset of chest CT images and simulated CT measurements with correlated noise and system blur. Our results show that Fourier diffusion models can improve image quality for supervised diffusion posterior sampling relative to existing conditional diffusion models. |
|---|---|
| ISSN: | 1558-254X |
| DOI: | 10.1109/TMI.2025.3553805 |