This repository contains code and data for the manuscript:
Classifying T cell activity in autofluorescence intensity images with convolutional neural networks. Zijie J Wang, Alex J Walsh, Melissa C Skala, Anthony Gitter. Journal of Biophotonics, 13:3, 2020.
Directory/File | Renderer | Description |
---|---|---|
images | Directory containing T-cell images | |
plots | Directory containing supplementary figures | |
resource | Directory containing supplementary training files and CellProfiler pipelines | |
image_processing.ipynb | Notebook for image cropping, filtering and mask measurement | |
frequency_classifier.ipynb | Notebook for a trivial frequency classifier | |
logistic_regression.ipynb | Notebook for fitting three logistic regression models with different feature vectors | |
simple_neural_network.ipynb | Notebook for training a one-layer fully connected neural network | |
simple_cnn_lenet.ipynb | Notebook for end-to-end training CNN LeNet | |
transfer_learning.ipynb | Notebook for fine-tuning layers of pre-trained Inception v3 |
- Nbviewer renders these notebooks as static HTML web pages.
- Binder hosts an executable environment for notebooks.
- Three
conda
environment files are provided.environment.yml
andenvironment-windows.yml
include main dependencies with major versions for macOS/Linux and Windows respectively;environment-complete.yml
provides a complete list of all packages and versions. - You can use
conda env create -f {environment.yml|environment-windows.yml|environment-complete.yml}
to install dependencies. - On Windows, the R package
ggpubr
must be installed separately after creating the conda environment with the commandRscript -e "install.packages('ggpubr')"
.
The software is available under the BSD 3-Clause Clear License.