Deep-learning based denoising and reconstruction of super-resolution structured illumination microscopy images
Problem
Noisy raw images in super-resolution structured illumination microscopy (SR-SIM), e.g. from low excitation power or short exposure, cause reconstruction artifacts.
Why it matters
High-quality super-resolution imaging of fluorescently labeled samples otherwise requires strong illumination, limiting what can be imaged.
What the AI did
A residual encoding-decoding convolutional neural network (RED-Net) denoised computationally reconstructed noisy SR-SIM images, and deep learning was also used to reconstruct raw SIM images directly into high-resolution images.
Results
not stated in source
Impact
Enables robust, artifact-resistant super-resolution reconstruction across various noise levels, per the abstract.
Record
| Venue | Photonics Research |
|---|---|
| Published | 2021 |
| Work type | article |
| Authors | Zafran Hussain Shah, Marcel W. Muller, Tung-Cheng Wang, Philip Maurice Scheidig, Axel Schneider, Mark Schüttpelz, Thomas R. Huser, Wolfram Schenck |
| Citations | 96 (OpenAlex, at import time) |
| References | 58 |
| Model or method | RED-Net (residual encoding-decoding convolutional neural network) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1364/prj.416437 |
| Open access | green |
| Keywords | deep learning, structured illumination microscopy, denoising, image reconstruction, RED-Net |
Importance rating
real but narrower advance
Narrower technical advance: denoising/reconstruction for SR-SIM microscopy, enabling imaging at lower illumination; no scale or breadth stated in abstract.