Quantitative-Microplankton-Tracker

By Harshith Bachimanchi, Benjamin Midtvedt, Daniel Midtvedt, Erik Selander, and Giovanni Volpe.

The quantitiative microplankton tracker is a deep-learning based framework to predict dry masses and three-dimensional positions of microplankton from experimental holographic images.

This repository contains source code and data for the article, Microplankton life histories revealed by holographic microscopy and deep learning

Description

The quantitative microplankton tracker uses two neural network in sequence to predict the dry masses and 3D positions of the microplanktons in holographic images. The first neural network, Regression U-Net (RU-Net), segments single plankton holograms from a large field-of-view holographic image and predicts the dry masses and vertical positions. The second neural network, Weighted Average Convolutional Neural network (WAC-Net), is then applied to the cropped image sequences of planktons to refine the plankton dry masses and vertical positions predicted by RU-Net.

Usage

Training-tutorials

We provide tutorials to train the neural networks, RU-Net, and, WAC-Net, from scratch in training-tutorials. PDF versions of training-tutorials are included for quick viewing.

1. 1-training_RU-Net.ipynb demonstrates the training process of RU-Net.

2. 2-training_WAC-Net_drymass.ipynb demonstrates the training process of WAC-Net to predict plankton dry masses.

3. 3-training_WAC-Net_vertical_positions.ipynb demonstrates the training process of WAC-Net to predict plankton vertical positions.

Usage examples

We also provide additional examples in examples on how to apply pre-trained neural networks on experimental holographic images. PDF versions of examples are included for quick viewing.

1. 1-example_RU-Net.ipynb demonstrates how to use RU-Net on experimental holographic images. The notebook also generates figure 1 and figure 2 of the paper.

2. 2-example_WAC-Net.ipynb demonstrates how to use WAC-Net on cropped image sequences of planktons to obtain a refined dry mass value. The notebook also generates figure 3 and figure 4 in the paper.

3. 3-example_Analysis.ipynb demonstrates the complete analysis pipeline, i.e., using RU-Net and WAC-Net in sequence.

Citation

If you use this code for your research, please cite our paper:

https://elifesciences.org/articles/79760

Harshith Bachimanchi, Benjamin Midtvedt, Daniel Midtvedt, Erik Selander, Giovanni Volpe (2022).
"Microplankton life histories revealed by holographic microscopy and deep learning." 
eLife 11:e79760.
https://doi.org/10.7554/eLife.79760

See also DeepTrack2.1:

Benjamin Midtvedt, Saga Helgadottir, Aykut Argun, Jesús Pineda, Daniel Midtvedt, Giovanni Volpe.
"Quantitative Digital Microscopy with Deep Learning."
Applied Physics Reviews 8 (2021), 011310.
https://doi.org/10.1063/5.0034891

Funding

This work was partly supported by the H2020 European Research Council (ERC) Starting Grant ComplexSwimmers (Grant No. 677511), the Horizon Europe ERC Consolidator Grant MAPEI (Grant No. 101001267), the Knut and Alice Wallenberg Foundation (Grant No. 2019.0079), and the Swedish Research Council (VR, Grant No. 2019-05238).