RSNA Intracranial Hemorrhage Detection - RSNA/AI-Challenge-Data GitHub Wiki
RSNA assembled this dataset in 2019 for the RSNA Intracranial Hemorrhage Detection AI Challenge (https://www.kaggle.com/c/rsna-intracranial-hemorrhage-detection/). De-identified head CT studies were provided by four research institutions. A group of over 60 volunteer expert radiologists recruited by RSNA and the American Society of Neuroradiology labeled over 25,000 exams for the presence and subtype classification of acute intracranial hemorrhage.
Description
The dataset is contained in a Zip archive that includes both DICOM image files (.dcm) and a tabular annotation file (.csv). A detailed description of the dataset is provided in Flanders AF et al. "Construction of a Machine Learning Dataset through Collaboration: The RSNA 2019 Brain CT Hemorrhage Challenge," Radiology: Artificial Intelligence 2020;2:3 (https://pubs.rsna.org/doi/10.1148/ryai.2020190211).
License
You may access and use these de-identified imaging datasets and annotations (“the data”) for non-commercial purposes only, including academic research and education, as long as you agree to abide by the following provisions: Not to make any attempt to identify or contact any individual(s) who may be the subjects of the data. If you share or re-distribute the data in any form, include a citation to the “Brain CT Hemorrhage Dataset, Copyright RSNA, 2019” as follows: Flanders AF, et al. The RSNA Brain CT Hemorrhage Dataset [10.1148/ryai.2020190211]. Radiology: Artificial Intelligence 2020;2:3."
Tutorial
Files
This was a two-stage challenge. You will need the images provided as stage_2_test.zip. You will also need the training data - stage_2_train.csv - and the sample submission stage_2_sample_submission.csv, which provides the IDs for the test set, as well as a sample of what a submission should look like.
Note: Also review two-stage FAQs for more details.
Data Format
The training data is provided as a set of image Ids and multiple labels, one for each of five sub-types of hemorrhage, plus an additional label for any, which should always be true if any of the sub-type labels is true.
There is also a target column, Label, indicating the probability of whether that type of hemorrhage exists in the indicated image.
There will be 6 rows per image Id. The label indicated by a particular row will look like [Image Id]_[Sub-type Name], as follows:
Id,Label
1_epidural_hemorrhage,0
1_intraparenchymal_hemorrhage,0
1_intraventricular_hemorrhage,0
1_subarachnoid_hemorrhage,0.6
1_subdural_hemorrhage,0
1_any,0.9
DICOM Images
All provided images are in DICOM format. DICOM images contain associated metadata. This will include PatientID, StudyInstanceUID, SeriesInstanceUID, and other features.
Prediction
In this challenge competitors predicted whether a hemorrhage exists in a given image, and what type it is.
File descriptions
stage_2_train.csv - the training set. Contains Ids and target information. stage_2_sample_submission.csv - a sample submission file in the correct format. Contains Ids for the test set.
Data fields
Id - An image Id. Each Id corresponds to a unique image, and will contain an underscore. Label - The probability of whether that sub-type of hemorrhage (or any hemorrhage in the case of any) exists in the indicated image.