Datasets:
annotations_creators:
- no-annotation
language_creators:
- other
language:
- en
license:
- other
multilinguality:
- monolingual
size_categories:
- 100B<n<1T
source_datasets:
- original
task_categories:
- image-classification
task_ids:
- multi-label-image-classification
pretty_name: ColonCancerCTDataset
tags:
- colon cancer
- medical
- cancer
dataset_info:
features:
- name: image
dtype: image
- name: ImageType
sequence: string
- name: StudyDate
dtype: string
- name: SeriesDate
dtype: string
- name: Manufacturer
dtype: string
- name: StudyDescription
dtype: string
- name: SeriesDescription
dtype: string
- name: PatientSex
dtype: string
- name: PatientAge
dtype: string
- name: PregnancyStatus
dtype: string
- name: BodyPartExamined
dtype: string
splits:
- name: train
num_bytes: 3537157
num_examples: 30
download_size: 3538117
dataset_size: 3537157
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
Dataset Card Creation Guide
Table of Contents
- Dataset Card Creation Guide
Dataset Description
- Homepage: https://portal.imaging.datacommons.cancer.gov
- Repository: https://aws.amazon.com/marketplace/pp/prodview-3bcx7vcebfi2i#resources
- Paper: https://aacrjournals.org/cancerres/article/81/16/4188/670283/NCI-Imaging-Data-CommonsNCI-Imaging-Data-Commons
Dataset Summary
The dataset in the focus of this project is a curated subset of the National Cancer Institute Imaging Data Commons (IDC), specifically highlighting CT Colonography images. This specialized dataset will encompass a targeted collection from the broader IDC repository hosted on the AWS Marketplace, which includes diverse cancer imaging data. The images included are sourced from clinical studies worldwide and encompass modalities such as Computed Tomography (CT), Magnetic Resonance Imaging (MRI), and Positron Emission Tomography (PET).
In addition to the clinical images, essential metadata that contains patient demographics (sex and pregnancy status) and detailed study descriptions are also included in this dataset, enabling nuanced analysis and interpretation of the imaging data.
Supported Tasks
The dataset can be utilized for several tasks:
- Developing machine learning models to differentiate between benign and malignant colonic lesions.
- Developing algorithms for Creating precise algorithms for segmenting polyps and other colonic structures.
- Conducting longitudinal studies on cancer progression.
- Assessing the diagnostic accuracy of CT Colonography compared to other imaging modalities in colorectal conditions.
Languages
English is used for text data like labels and imaging study descriptions.
Dataset Structure
Data Instances
The data will follow the structure below:
''' { "image": image.png # A CT image, "ImageType": ['ORIGINAL', 'PRIMARY', 'AXIAL', 'CT_SOM5 SPI'] # A list containing the info of the image, "StudyDate": "20000101" # Date of the case study, "SeriesDate": 20000101" # Date of the series, "Manufacturer": "SIEMENS" # Manufacturer of the device used for imaging, "StudyDescription": "Abdomen^24ACRIN_Colo_IRB2415-04 (Adult)" # Description of the study, "SeriesDescription": "Colo_prone 1.0 B30f" # Description of the series, "PatientSex": "F" # Patient's sex, "PatientAge": "059Y" # Patient's age, "PregnancyStatus": "None" # Patient's pregnancy status, "BodyPartExamined": "COLON" # Body part examined } '''
Data Fields
- image (PIL.PngImagePlugin.PngImageFile): The CT image in PNG format
- ImageType (List(String)): A list containing the info of the image
- StudyDate (String): Date of the case study
- SeriesDate (String): Date of the series study
- Manufacturer (String): Manufacturer of the device used for imaging
- StudyDescription (String): Description of the study
- SeriesDescription (String): Description of the series
- PatientSex (String): Patient's sex
- PatientAge (String): Patient's age
- PregnancyStatus (String): Patient's pregnancy status
- BodyPartExamined (String): The body part examined
Data Splits
train | validation | test | |
---|---|---|---|
Average Sentence Length |
Dataset Creation
Curation Rationale
The dataset is conceived from the necessity to streamline a vast collection of heterogeneous cancer imaging data to facilitate focused research on colon cancer. By distilling the dataset to specifically include CT Colonography, it addresses the challenge of data accessibility for researchers and healthcare professionals interested in colon cancer. This refinement simplifies the task of obtaining relevant data for developing diagnostic models and potentially improving patient outcomes through early detection. The curation of this focused dataset aims to make data more open and usable for specialists and academics in the field of colon cancer research.
Source Data
According to IDC, data are submitted from NCI-funded driving projects and other special selected projects.
Personal and Sensitive Information
According to IDC, submitters of data to IDC must ensure that the data have been de-identified for protected health information (PHI).
Considerations for Using the Data
Social Impact of Dataset
The dataset tailored for CT Colonography aims to enhance medical research and potentially aid in early detection and treatment of colon cancer. Providing high-quality imaging data empowers the development of diagnostic AI tools, contributing to improved patient care and outcomes. This can have a profound social impact, as timely diagnosis is crucial in treating cancer effectively.
Discussion of Biases
Given the dataset's focus on CT Colonography, biases may arise from the population demographics represented or the prevalence of certain conditions within the dataset. It is crucial to ensure that the dataset includes diverse cases to mitigate biases in model development and to ensure that AI tools developed using this data are generalizable and equitable in their application.
Other Known Limitations
The dataset may have limitations in terms of variability and scope, as it focuses solely on CT Colonography. Other modalities and cancer types are not represented, which could limit the breadth of research.
Licensing Information
https://fairsharing.org/FAIRsharing.0b5a1d
Citation Information
Provide the BibTex-formatted reference for the dataset. For example:
@article{fedorov2021nci,
title={NCI imaging data commons},
author={Fedorov, Andrey and Longabaugh, William JR and Pot, David
and Clunie, David A and Pieper, Steve and Aerts, Hugo JWL and
Homeyer, Andr{\'e} and Lewis, Rob and Akbarzadeh, Afshin and
Bontempi, Dennis and others},
journal={Cancer research},
volume={81},
number={16},
pages={4188--4193},
year={2021},
publisher={AACR}
}