Fluorescence microscopy datasets for training deep neural networks
Problem
Deep learning methods that improve fluorescence microscopy signal-to-noise (allowing shorter exposures and less photobleaching/phototoxicity) require large paired training datasets that were lacking.
Why it matters
Photobleaching and light-induced phototoxicity limit live-cell fluorescence imaging in biological research.
What the AI did
The paper provides high-quality paired long/short-exposure fluorescence microscopy datasets for training and evaluating deep learning methods; it does not itself train or demonstrate a model.
Results
Provides high-quality paired long/short-exposure datasets usable to train and evaluate deep learning methods; no performance metrics stated
Impact
Enables development of deep learning methods to reduce photobleaching and phototoxicity; no demonstrated outcome in this evidence.
Record
| Venue | GigaScience |
|---|---|
| Published | 2021 |
| Work type | article |
| Authors | Guy M. Hagen, Justin Bendesky, Rosa Machado, Tram-Anh Nguyen, Tanmay Kumar, Jonathan Ventura |
| Citations | 32 (OpenAlex, at import time) |
| References | 46 |
| Model or method | convolutional neural networks (generic, no specific model named) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1093/gigascience/giab032 |
| Open access | gold |
| Keywords | fluorescence microscopy, deep learning, datasets, photobleaching, phototoxicity, convolutional neural networks |
Importance rating
real but narrower advance
Photobleaching/phototoxicity limit live-cell fluorescence imaging across many areas of biological research; this is a foundational research tool, but the paper only provides training datasets, a narrower advance.