Threshold independent detection and localization of diffraction-limited spots.
In biomedical microscopy data, a common task involves the detection of diffraction-limited spots that visualize single proteins, domains, mRNAs, and many more. These spots were traditionally detected with mathematical operators such as Laplacian of Gaussian. These operators, however, rely on human input ranging from image-intensity thresholds, approximative spot sizes, etc. This process is tedious and not always reliable. DeepBlink relies on neural networks to automatically find spots without the need for human intervention. DeepBlink is available as a ready-to-use command-line interface.
| Usage | Example |
|---|---|
 |
 |
More documentation about deepBlink including how to train, create a dataset, contribute etc. is available at https://github.com/BBQuercus/deepBlink/wiki.
This package is built for Python versions newer than 3.6 and can easily be installed with pip:
pip install deepblinkOr using conda:
conda install -c bbquercus deepblinkAdditionally for GPU support, install tensorflow-gpu through pip and with the
appropriate CUDA and cuDNN verions matching your GPU setup. Lastly, you can also use our KNIME node for inference. Please follow the installation instructions on KNIME hub.
A video overview can be found here. Inferencing on deepBlink is performed at the command line as follows:
deepblink predict -m MODEL -i INPUT [-o OUTPUT] [-r RADIUS] [-s SHAPE]With MODEL being a pre-trained or custom model and INPUT being the path to a input image or folder containing images.
deepBlink is currently available on Nucleic Acid Research here. If you find deepBlink useful, consider citing us:
@article{10.1093/nar/gkab546,
author = {Eichenberger, Bastian Th and Zhan, YinXiu and Rempfler, Markus and Giorgetti, Luca and Chao, Jeffrey A},
title = "{deepBlink: threshold-independent detection and localization of diffraction-limited spots}",
journal = {Nucleic Acids Research},
year = {2021},
month = {07},
issn = {0305-1048},
doi = {10.1093/nar/gkab546},
url = {https://doi.org/10.1093/nar/gkab546},
note = {gkab546},
eprint = {https://academic.oup.com/nar/advance-article-pdf/doi/10.1093/nar/gkab546/38848972/gkab546.pdf},
}