Cell Nuclei EBB Dataset

Dataset description

Cell Nuclei EBB Dataset is comprised of 17,455 images, captured in a Sartorius IncuCyte Incubator/Microscope, with 20x magnification.

This dataset was generated in order to train a deep learning model for nuclei detection in brightfield images (unpublished) at LabSinal laboratory (https://www.ufrgs.br/labsinal/), and contains annotations on 387,245 cancer/fibroblast cells' nuclei.

Annotations are described as elipses (Eliptical Bounding Boxes - EBB) by columns: img_file_name,cx,cy,width,height,angle, stored in "annotations.csv" file, in which:

img_file_name: refers to the source image (one of the 17,455 present in the "imgs" folder);cx: pixel coordinate of the X center of the current EBB;cy: pixel coordinate of the Y center of the current EBB;width: lateral dimension, in pixels, of the current EBB;height: lateral dimension, in pixels, of the current EBB;angle: angle, in degrees, relative to the horizontal, of the current EBB.The annotation process was performed using forNMA (unpublished), a tool for the automatic nucleus segmentation in fluorescence images. All images in this dataset had its respective fluorescence channel image (data not shown), whose fluorescent marker (either H2B, 53BP1 or DAPI) was used to obtain nucleus coordinates/shape.

Further information

Cell lines (as described in column "cell_line" in "lines_and_treatments_info.csv" file):, and also in "dataset_description.csv" file):

A172, U87 and U251: human glioblastomaMRC5: human lung fibroblastCultivation condition (as described in column "treatment" in "lines_and_treatments_info.csv", and also in "dataset_description.csv" file):

CTR: cells belonging to the control group - without the addition of chemotherapy;TMZ: Temozolomide;ACT: Actinomycin D;EGF: Epidermal Growth Factor;3MA: 3-Methyladenine;HCQ: Hydroxychloroquine;TMZ_ACT: Combination of Temozolomide and Actinomycin D;TRAM: Trametinib.

Additional information files are stored in csv files, inside main folder.

Annotations as previously described in the Abstract section are stored in "annotations.csv" file.

Example image

This example image illustrates the annotations (performed by forNMA, shown in green) compared to the trained model (retrained R2CNN, shown in blue).