Reference Information
=====================

Dataset Attribution and Usage
-----------------------------

* Dataset Title: "Segmentations of carcinogenic DNA damages in WSI"

* Persistent Identifier: https://doi.org/10.5281/zenodo.7266845

* Dataset Contributors: Wagner, Patrick; Springenberg, Maximilian; Kröger, Marius,; Moritz, Rose K. C.; Schleusener, Johannes; Meinke, Martina C.; Ma, Jackie

* License: Use of these data is covered by the following license:
  * Title: 
  * Specification: 

* Suggested Citations:

  * Dataset citation:
    > Wagner, Patrick; Springenberg, Maximilian; Kröger, Marius,; Moritz, Rose K. C.; Schleusener, Johannes; Meinke, Martina C.; Ma, Jackie. 2022. Data for the article 	"Semantic modeling of cell damage prediction: A machine learning approach at human-level performance in dermatology",Zenodo, Dataset, 		https://doi.org/10.5281/zenodo.7266845

Additional Dataset Metadata
===========================

Dates and Locations
-------------------

* Dates of data collection: Images were collected between June 2020 and August 2022.

Methodological Information
==========================

* Methods of data collection/generation: see manuscript for details

- - -

Data and File Overview
======================

Summary Metrics
---------------

* File count: 805
* File formats: .png(804); .csv(1)
* File dimensions: 1384x1040 pixel
* Bit depth: 24 bit (files ending on _highdamage.png and _mask.png have a bitdepth of 8 bit)
* Description: WSI images of the carcinogenic DNA damages caused by ultraviolet (UV) radiation on excised abdominal human skin, excised porcine skin and in vitro skin models with different melanin indices


Naming Conventions
------------------

* File naming scheme of png files: The files are named as follows. Please refer to the manuscript about the protocols for the image acquisition

* Images: type_date-#id staining magnification.png

* epidermis masks: type_date-#id staining magnification_mask.png

* high_damage masks: type_date-#id staining magnification_highdamage.png


		*type: HH = excised human skin; SWO = excised porcine skin; RHE = reconstructed skin model; MAA = skin model after african american skin; MAC = skin model after asian caucasian skin
		*date: date the skin was radiated with UV light
		*#id: running number of slices; neg ctrl = negative staining control; pos ctrl = positive staining control
		*staining: CPD: slice stained for cyclobutane pyrimidine dimers DNA damages; 6-4PP: slice stained for pyrimidine (6-4) pyrimidone photoproducts DNA damages
		*magnifiction: number in mW (between 2mW-50mW)

File/Folder Details
===================

Details for: database.csv
---------------------------------------

* Description: a comma-delimited file containing the DNA damage scores for all WSI

* Format(s): .csv

* Size(s): 76.0 KB

* Dimensions: 805 rows x 9 columns

* Variables:
  * path_image: Path to the input image.
  * path_epidermis: Path to the annotated epidermis segmentation maps (maps are inverted).
  * path_high_famage: Path to the annotated high-damage segmentation maps (maps are inverted).
  * score_nuclei: Score, describing the relative amount of damaged nuclei (bounded in [0,1]).
  * score_area: Score, describing the relative amount of damaged epidermis area (bounded in [0,1]).
  * source_tissue: Type of tissue. Humanhaut = excised human skin; Hautmodell = skin model; Schweineohr = excised porcine skin
  * patient_id: An ID, assigning samples to a patient (anonymized).
  * strat_fold: Stratified crossvalidation-fold ID.