Phase contrast images of bacteria and ground truth segmentations

Name: Phase contrast images of bacteria 

Data type: Paired microscopy images and corresponding labels/masks used for model training, organized as recommended by the DenoiSeg documentation.

Microscopy data type: Light microscopy (Phase Contrast)

Manual annotations: Labels/masks obtained via manual segmentation. For each region, all cells were annotated manually. Uncertain objects were left unannotated.

Microscope: Zeiss Axio Imager M2 epi-fluorescence microscope with a Zeiss Plan-Apochromat; 100x/1.4 oil DIC objective

File format: .tif (float 32-bits for phase contrast and 16-bit for mask images)

Image size: 256x256 pixels (Pixel size: 64.5 nm)

Content: 

train - raw (33 files) 

      - masks (33 files)

test - raw (11 files) 

     - masks (11 files)

All images available in the raw folders were normalized by dividing the original images with a gaussian blurred version or the original image (200 pixels). A groovy code working within ImageJ/Fiji corresponding to this operation is as follow:

ImagePlus normalize(ImagePlus input_image) {
    flatfield = (new Duplicator()).run(input_image)
    (new GaussianBlur()).blur(flatfield.getProcessor(), 200)
    return ImageCalculator.run(input_image, flatfield, "Divide create 32-bit")
}

import ij.ImagePlus
import ij.plugin.Duplicator
import ij.plugin.ImageCalculator
import ij.plugin.filter.GaussianBlur