TrackSegNet

Trajectory segmentation into diffusive states using LSTM neural network.

Installation and requirements

Clone the repository

git clone https://github.com/hkabbech/TrackSegNet.git
cd TrackSegNet

Activate the python environment

On Unix or MacOS, run:

source tracksegnet-env/bin/activate

On Windows, run:

tracksegnet-env\bin\activate

Note, to deactivate the virtual environment, type deactivate

Or Create the environment and install the packages

pip3 install virtualenv
virtualenv -p /usr/bin/python3 tracksegnet-env
source tracksegnet-env/bin/activate # for Windows: tracksegnet-env\bin\activate
python -m pip install -r requirements.txt

Prepare your data

Data organization

Organize your data in a folder SPT_experiment, each sub-folder should contain a file storing the trajectory coordinates in a MDF or CSV file format.

If CSV format is used, the headers should be: x, y, frame, track_id

.
├── data/
│   └── SPT_experiment/
│       ├── Cell_1
│       │    ├── *.tif
│       │    └── *.mdf
│       ├── Cell_2
│       │    ├── *.tif
│       │    └── *.mdf
│       ├── Cell_3
│       │    ├── *.tif
│       │    └── *.mdf
│       └── ...
│
├── src/
├── tracksegnet-env/
├── parms.csv
├── tracksegnet.py
└── ...

Change the main parameters

Update the main parameters in the parms.csv file according to your experiment:

• data_path: the path containing your data folder SPT_experiment to analyze
• track_format: The format of the files containing the trajectory coordinates, should be MDF or CSV
• time_frame: the time interval between two trajectory points (in second)
• pixel_size: the dimension of a pixel (in µm)
• num_states: the number of diffusive states for the classification(from 2 to 6 states)
• state_X_diff: The expected diffusion value for state X (in µm^2/s).
• state_X_alpha: The expected anomalous exponent α value for state X (from 0 to 2 -- ]0-1[: subdiffusion, 1: Brownian motion, ]1-2[: superdiffusion).
• pt_i_j: the probability of transitionning from the state i to the state j. The total number of probabilities should be $N^2$.

Note that the program will run on the toy example if the parameters are unchanged.

For updating the parameters of the track simulation and neural network training, please make the changes in the main file tracksegnet.py.

Run the program

./tracksegnet.py parms.csv

Reference

Arts, M., Smal, I., Paul, M. W., Wyman, C., & Meijering, E. (2019). Particle Mobility Analysis Using Deep Learning and the Moment Scaling Spectrum. Scientific Reports, 9(1), 17160. https://doi.org/10.1038/s41598-019-53663-8.