Script usage

If you want to process your images directly without notebooks, there are two example pipelines to apply to images with stacks of channels, or with each channel in a different file.

In this case, you will not be able to interactively check which thresholdings and percentiles best apply to each channel.

The scripts apply the pipeline:

• saturation of outliers

• channel normalization

• thresholding

• percentile filtering

• save

The following code is only a snapshot, please check the full script.

For stacks of channels, and with all the parameters defined, the general idea would be as follow:

images_original = list(map(IP.parse_image_pages, files))

imgs_out = map(lambda p: IPrep.remove_outliers(p, up_limit, down_limit), images_original)
imgs_norm = map(IPrep.normalize_channel_cv2_minmax, imgs_out)
filtered_images = map(lambda i: preprocess_image(i, thresholds, percentiles), imgs_norm)
imgs_filtered = list(filtered_images)

# save with channel names
images_final = map(
    lambda p, f: IPrep.save_img_ch_names_pages(p, f, ch_last=True, channel_names=channel_names),
    imgs_filtered, names_save)

preprocess_image is a function defined in the example and applies thresholding and percentile per channel.

For channels defined by a single file and organized in patient folders, and with all the parameters defined, the general idea would be as follow:

for channel, th, perc in zip(channel_names, thresholds, percentiles):
    file_paths = [file for file in files if str(channel + '.ome.tiff') in str(file)]
    images_original = list(map(IP.parse_image, file_paths))
    imgs_out = map(lambda p: IPrep.remove_outliers(p, up_limit, down_limit), images_original)
    imgs_norm = map(IPrep.normalize_channel_cv2_minmax, imgs_out)
    if isinstance(threshold, float):
        imgs_filtered = list(map(lambda p: IPrep.out_ratio2(p, th=threshold), imgs_norm))
    if percentile is not None:
        imgs_filtered = map(
            lambda p: IPrep.percentile_filter(p, window_size=3, percentile=percentile, transf_bool=True),
            imgs_filtered)

    map(lambda p, f: IPrep.save_images(p, f, ch_last=True), imgs_filtered, names_save)

Please check the scripts for additional parameters. Feel free to adjust all the code.