Toward reliable fluorescence imaging: Optical prior-guided probabilistic reconstruction for structured illumination microscopy
Fluorescence microscopy is expected to be a reliable technique for bioprocess analysis, but the current single, deterministic imaging cannot objectively reflect the inherent observation errors caused by instruments and algorithms. For structured illumination microscopy (SIM) used for fast and long-t...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
AIP Publishing LLC
2025-06-01
|
| Series: | APL Photonics |
| Online Access: | http://dx.doi.org/10.1063/5.0270116 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1839642025679388672 |
|---|---|
| author | Kun Lin Junkang Dai Huaian Chen Yi Jin |
| author_facet | Kun Lin Junkang Dai Huaian Chen Yi Jin |
| author_sort | Kun Lin |
| collection | DOAJ |
| description | Fluorescence microscopy is expected to be a reliable technique for bioprocess analysis, but the current single, deterministic imaging cannot objectively reflect the inherent observation errors caused by instruments and algorithms. For structured illumination microscopy (SIM) used for fast and long-term imaging at low excitation levels, the risk of unreliable misconceptions will be more non-negligible due to severe noise and super-resolution reconstruction. Here we present PG-SIM, a probabilistic SIM reconstruction method based on Bayesian neural networks and incorporating graph representation learning (GRL) to model optical prior knowledge. PG-SIM provides uncertainty and confidence maps corresponding to the imaging results, allowing biologists to simultaneously and quantitatively identify potential imaging errors without any reference. Furthermore, by leveraging the strong cognition ability of GRL to precisely learn the hierarchical representations of the SIM imaging process, PG-SIM itself also achieves a significant uncertainty reduction compared to current methods, overcoming the constraints of unreliability on the practical application of SIM. We believe that this work can inspire the development of more utilitarian and rigorous SIM techniques in the future. |
| format | Article |
| id | doaj-art-2ea4c64c7b0340b7b49e8eb0aec77fc8 |
| institution | Matheson Library |
| issn | 2378-0967 |
| language | English |
| publishDate | 2025-06-01 |
| publisher | AIP Publishing LLC |
| record_format | Article |
| series | APL Photonics |
| spelling | doaj-art-2ea4c64c7b0340b7b49e8eb0aec77fc82025-07-02T17:40:04ZengAIP Publishing LLCAPL Photonics2378-09672025-06-01106066108066108-1610.1063/5.0270116Toward reliable fluorescence imaging: Optical prior-guided probabilistic reconstruction for structured illumination microscopyKun Lin0Junkang Dai1Huaian Chen2Yi Jin3Department of Precision Machinery and Precision Instruments, University of Science and Technology of China, Hefei 230022, ChinaDepartment of Precision Machinery and Precision Instruments, University of Science and Technology of China, Hefei 230022, ChinaDepartment of Precision Machinery and Precision Instruments, University of Science and Technology of China, Hefei 230022, ChinaDepartment of Precision Machinery and Precision Instruments, University of Science and Technology of China, Hefei 230022, ChinaFluorescence microscopy is expected to be a reliable technique for bioprocess analysis, but the current single, deterministic imaging cannot objectively reflect the inherent observation errors caused by instruments and algorithms. For structured illumination microscopy (SIM) used for fast and long-term imaging at low excitation levels, the risk of unreliable misconceptions will be more non-negligible due to severe noise and super-resolution reconstruction. Here we present PG-SIM, a probabilistic SIM reconstruction method based on Bayesian neural networks and incorporating graph representation learning (GRL) to model optical prior knowledge. PG-SIM provides uncertainty and confidence maps corresponding to the imaging results, allowing biologists to simultaneously and quantitatively identify potential imaging errors without any reference. Furthermore, by leveraging the strong cognition ability of GRL to precisely learn the hierarchical representations of the SIM imaging process, PG-SIM itself also achieves a significant uncertainty reduction compared to current methods, overcoming the constraints of unreliability on the practical application of SIM. We believe that this work can inspire the development of more utilitarian and rigorous SIM techniques in the future.http://dx.doi.org/10.1063/5.0270116 |
| spellingShingle | Kun Lin Junkang Dai Huaian Chen Yi Jin Toward reliable fluorescence imaging: Optical prior-guided probabilistic reconstruction for structured illumination microscopy APL Photonics |
| title | Toward reliable fluorescence imaging: Optical prior-guided probabilistic reconstruction for structured illumination microscopy |
| title_full | Toward reliable fluorescence imaging: Optical prior-guided probabilistic reconstruction for structured illumination microscopy |
| title_fullStr | Toward reliable fluorescence imaging: Optical prior-guided probabilistic reconstruction for structured illumination microscopy |
| title_full_unstemmed | Toward reliable fluorescence imaging: Optical prior-guided probabilistic reconstruction for structured illumination microscopy |
| title_short | Toward reliable fluorescence imaging: Optical prior-guided probabilistic reconstruction for structured illumination microscopy |
| title_sort | toward reliable fluorescence imaging optical prior guided probabilistic reconstruction for structured illumination microscopy |
| url | http://dx.doi.org/10.1063/5.0270116 |
| work_keys_str_mv | AT kunlin towardreliablefluorescenceimagingopticalpriorguidedprobabilisticreconstructionforstructuredilluminationmicroscopy AT junkangdai towardreliablefluorescenceimagingopticalpriorguidedprobabilisticreconstructionforstructuredilluminationmicroscopy AT huaianchen towardreliablefluorescenceimagingopticalpriorguidedprobabilisticreconstructionforstructuredilluminationmicroscopy AT yijin towardreliablefluorescenceimagingopticalpriorguidedprobabilisticreconstructionforstructuredilluminationmicroscopy |