Wiki

With the Quantius Job folder as MATLAB's working directory, prepare images for mixed mode.

#!matlab

organoids.segment_nuclei_missing_slices.prepare_images

This will make the folders needed for storing the raw and processed images, as well as set up which images to use for training and testing. If there are a lot of images (1,000+), this may take a few minutes to run.

Next, train models on the slices with Quantius results and use those models to make predictions on the remaining slices.

#!matlab

organoids.segment_nuclei_missing_slices.train_and_test_model

This will actually train the models and use them to predict nuclear segmentations for all nuclear slices. Some morphological operations are used to convert the raw output into "objects". Though parallel processing is used, this can still take ~12 hours to run for ~2,000 images on my desktop iMac.