DISCO Cell Localization CNN Scripts

This repository contains the relevant files for generation and visualization of the convolutional neural network (CNN) presented in the forthcoming publication indicated at the bottom of this page.

Scripts

• kfold_hyperopt_binary_localizer.py - Uses the hyperopt package to explore and optimize the hyperparameter space of the CNN.
• optimized_binary_localizer.py - Trains the final network with the optimized (and manually selected) hyperparameters.

Notebooks

• genetic_optimization_visualizer.ipynb - Shows the results of the 'kfold_hyperopt_binary_localizer.py' script for selection of the optimal network hyperparameters.
• training_log_plotter.ipynb - Plots the training history of the optimized network.
• output_visualizer.ipynb - Loads the final network and shows segmentation predictions on validation data as well as the necessary downstream processing for laser vector path generation.

Data

• binary_data.p - The cell image data and annotations used as input to the network (training and validation).
• genetic_trials_cv.p - Output from 'kfold_hyperopt_binary_localizer.py' showing the optimization trials.
• binary_localizer_16_0.28892_1_54_7_12.log - The training log produced by 'optimized_binary_localizer.py'
• binary_localizer_16_0.28892_1_54_7_12.hdf5 - The final network used for automated cell laser lysis.

Note these scripts were tested on:

• tensorflow==1.13.1
• hyperopt==0.2.3
• keras==2.2.4
• numpy==1.16.2
• opencv-python==3.3.0
• sklearn==0.20.1

Publication:

Lamanna, J.* , Scott, E.Y. * , Edwards, H.* , Chamberlain, M.D., Dryden, M.D.M., Peng, J., Mair, B., Lee, A., Sklavounos, A.A., Abbas, F., Moffat, J. & A.R. Wheeler. "Digital Microfluidic Isolation of Single Cells for - Omics". 2020.
* Co-first authors DOI: (TBD)