I maintain this list mostly as a personal braindump of interesting medical datasets, with a focus on medical imaging.
Rather than try to group / cluster datasets, I'm going to try to maintain a set of keywords for each.
See commit log for a list of additions over time.
Please feel free to contribute!
Disclaimer: please remember to solve real clinical problems ☺
414 T1 MRIs from the OASIS dataset, processed using FreeSurfer and SAMSEG
Includes original images, along with processed volumes and resulting anatomical segmentation maps
Keywords: large, MRI, segmentations, labels, annotations, processed
Full-head images and ground-truth brain masks from 622 MRI, CT, and PET scans
Includes a landscape or MRI scans with different contrasts, resolutions, and populations from infants to glioblastoma patients
Also includes anatomical segmentation maps for a subset of the images
Keywords: large, diverse, multi-modal, brain masks, segmentations, brain extraction, skull stripping
224,316 chest radiographs of 65,240 patients, with labels from reports
Keywords: very-large, X-ray, labels
100000 radiographs
Keywords: very-large, X-ray, labels
371,920 chest x-rays associated with 227,943 imaging studies
3/16/2019: Not yet linked with MIMIC ICU data. See news article
v2: free-text radiology reports
Need to request access
Keywords: very-large, X-ray, labels
160,000 images from 67,000 patients that were interpreted and reported by radiologists
labeled with 174 different radiographic findings, 19 differential diagnoses and 104 anatomic locations organized as a hierarchical taxonomy mapped to standard Unified Medical Language System (UMLS)
Keywords: very-large, X-ray, labels
1000+ dataset of eye gaze, radiological reports, dictation, segmentation on MICMIC-CXR Database
code to reproduce experiments
Keywords: medium, X-ray, labels
Several collections
Tons of Images of various kinds, including CT, MR, Pathology, PT, with diagnoses
Keywords: vary-large, CT, MR, labels
Part of Cancer Imaging Archive
50000+ patients with CT data, some pathology, limited availability
Keywords: very-large, CT, labels
32000+ CT scans with annotations, meta-data, semantic labels from radiological reports
Keywords: very-large, CT, labels
10,000+ labeled echocardiogram videos and human expert tracing
Keywords: very-large, ultrasound, labels
MRI for 8500 young (9-10yo) subjects (about 4100 for training)
Keywords: large, MRI
4,000 simulated sinogram/image pairs of 2D breast CTs Keywords: large, CT, reconstruction
two large scale neuroimaging datasets on reading and language development
Over 3000 MRI, fMRI
article | more resources
Keywords: large, MRI
1,370 knee MRI exams with diagonsis (healthy/ACL tear/meniscal tear)
Keywords: large, MRI, labels
k-space data
1500 fully sample knee MRIs and 10K clinical MRIs, and 6.5K brain MRIs.
Part of a challenge
Keywords: large, MRI, k-space
Open-Access Multi-Coil k-Space Dataset for Cardiovascular Magnetic Resonance Imaging
k-space data, roughly 250 volumes
Keywords: medium, MRI, k-space
1704 MRI, 556 amyloid and tau CSF samples, blood markers, genetic info and longitudinal cognitive data on ~400 at risk individuals
Keywords: medium, MRI, genetics, labels
10 Medical image datasets with segmentations
2000+ CT & MR images of various organs from different sources
Keywords: medium, MRI, segmentations
Multiple Acquisitions for Standardization of Structural Imaging Validation and Evaluation
8000 diffusion-weighted volumes
10 3D FLAIR, T1-, and T2-weighted datasets of a single healthy subject
Keywords: large, MRI
1000+ fMRI and other modalities subjects with annotated event files; raw and preprocessed
Keywords: medium, fMRI
List of mri k-space datasets
601 series of CT projection data, reconstructed images, and clinical data reports Keywords: medium, CT, reconstruction
Brain MRI images together with manual FLAIR abnormality segmentation masks
110 subjects from TCIA LGG collection with lower-grade glioma cases
Keywords: medium, brain, MRI, segmentation, LGG, FLAIR
Few subjects, but many modalities (T1,T2,SWI,Angio,DWI, fMRI during Forrest Gump at 3T (audio+visual+eyetracking+physio) and 7T (audio+physio only), some audio tasks, and other important visual tasks)
Keywords: small, multi-modal
LIDC-IDRI consists of diagonstic and lung cancer screening CTs.
1018 cases with some Radiologist Annotations/Segmentations and nodule counts
Also available through LUng Nodule Analysis (LUNA) challenge
Keywords: large, CT, labels
922 breast cancer patients publicly available for machine learning and clinical research.
Contains breast MRIs, clinical, demographics, pathology, treatment, outcomes, and genomic data as well as image annotations (locations) and features.
Keywords: large, MRI, labels
All imaging
Fundus imaging
Keywords: very-large
high resolution tissue-scale volume electron microscopy (vEM) datasets acquired with the enhanced focused ion beam scanning electron microscopy (FIB-SEM) technology developed at Janelia. Accompanying these EM volumes are automated segmentations and analyses of intracellular sub-structures.
Keywords: very large, EM, segmentation
4703 CXR of COVID19 patients, manually annotated Brixia score
Keywords: large, x-ray, covid
349 CT images collected from several COVID19-related papers
Image captions
Keywords: medium, CT, covid
~5000 xrays
Keywords: medium, x-ray, pneumonia
998 Chest x-ray examinations from 361 COVID+ patients. Annotations with appearance classification and Airspace Disease Grading Clinical variables Keywords: large, x-ray, covid
1350+ Xrays, 150+ CTs, 800 diagnoses
Keywords: medium, CT, covid
1000+ CTs of COVID19 patients
50 are annotated per pixel
Keywords: large, CT, covid, segmentations
~250 chest CTs with positive RT-PCR SARS-CoV-2, annotations of COVID-19 lesions Keywords: medium, CT, covid, annotations, segmentations
~100 segmented CT slices
Keywords: medium, CT, segmentations, covid
~150 xrays, ongoing, some hospital data
Keywords: medium, x-ray, covid
ongoing, about 60 patients at last check, CT
paper pdf
Keywords: medium, CT, covid
345 subjects, 891 functional scans, and 27 diverse stories of varying duration totaling ~4.6 hours of unique stimuli (~43,000 words).
Nature paper
Keywords: medium, fMRI
1000 X-rays and 240 CTs with annotations (paper)
Keywords: large, CT, covid, segmentations
129 retinal images.
Keywords: small, fundus
40 retinal images with segmentations
Keywords: small, retinal, segmentations
500+ CT scans from 11+ countries with Abdominal Organ Segmentation (the liver, kidney, spleen, and pancreas)
Keywords: large, abdominal, CT
Various imaging (longitudinal MRI), Genetics, Clinical data
Several thousand patients
Keyworks: large, MRI, genetics, clinical
~120 image volumes (whole body CT and MRI images)
more than 1900 annotated anatomical structures
Keywords: medium, MRI, CT, whole-body, manual-segmentation
Seems like 101 manually labelled brain MRIs
Keywords: medium, MRI, brain, manual-segmentation
3000 brain scans (T1w, bold, events)
Standardized tests, scores, demographics
Keywords: large, MRI, fMRI, tests
A curated dataset of digital breast tomosynthesis images from 5,060 patients.
Keywords: large, tomosynthesis, DBT, breast, detection
2600+ scanned film mammography studies
Keywords: large, x-ray
63 manually labelled brain scans.
Costs ($1500?)
Discussion
Keywords: medium, MRI, brain, manual-segmentation, costly
This is a challenge for ISBI2019
22 particiapnts with cognitive and physiological mreasures, and 7T rs-fMRI
200+ subjects across several datasets (CTs, Xrays, MRIs)
20 cardiac MR images in Congenital Heart Disease
paper
~50 children (~10yo) with single follow-up with MRI, fMRI and assesments
Keywords: medium, fMRI, longitudinal
paper
3T fMRI 132 typical dev children, 2 time points, four tasks
Keywords: medium, fMRI, longitudinal
aggregates auditory story-listening fMRI datasets acquired over the course of roughly seven years
Keywords: medium, fMRI
229 T1-weighted MRI scans (n=220) with lesion segmentation
MNI152 standard-space T1-weighted average structural template image
A .csv file containing lesion metadata
paper
Keywords: medium, MRI, segmentations
21 Canine mammary carcinoma whole slide images.
Annotated by 2/3 experts
Keywords: small, 2D, whole slide imaging
48 manually annotated in utero fetal MR
Keywords: small, mri, fetal, labels
Single voluneer, 73 Sessions at multiple sites over ~17 years
MRI, at least T1 at each session, with other modalities varying by session.
Phenotype file provided
Keywords: small, MRI, longitudinal
Single volume, histological space , 100 micron) with GM/WM surfaces and cortical layers
ftp://bigbrain.loris.ca | interactive
Keywords: small, histology, high-resolution, segmentations
Single volume, ultra-high resolution MRI dataset (100-micron)
Keywords: small, MRI, brain
8-subjects large-scale fMRI (40-sessions, high sampling, high resolution). T1w, T2w, T2*w MRI
Video description
Keywords: small, MRI, brain, fMRI
(ex-vivo) brain MRIs or brains of different animals
Keywords: small, MRI, brain, animals
Three Diffusion of healthy traveling adults
Keywords: small, MRI, diffusion, brain
Prenatal brain MRI samples (looks like single subject?)
Keywords: small, MRI, fetal
1058 wholes slide images (WSIs) with corresponding clinical characteristics
Part of tumor regions are annotated in WSIs.
Clinical characteristics include age, tumor size, tumor type, ER, PR, HER2, HER2 expression, histological grading, surgical, Ki67, molecular subtype, number of lymph node metastases, and ALN status
Paper reference
Keywords: large, breast cancer, multi-modal, WSI, clinical characteristics
4870 registered HE-IHC image pairs, covering four expression levels of HER2 (0, 1+, 2+, 3+).
Keywords: large, breast cancer, HE, IHC
This dataset is curated based on MIMIC-CXR, containing 3 metadata files that consist of pulmonary edema severity grades extracted from the MIMIC-CXR dataset through different means: 1) by regular expression (regex) from radiology reports, 2) by expert labeling from radiology reports, and 3) by consensus labeling from chest radiographs.
Keywords: pulmonary edema, severity grades, chest x-ray, radiology reports, MIMIC-CXR
predict sepsis in an ICU population
5000 ICU patients in three separate hospital systems
detailed information about critical care stays for over 200,000 admissions at 200+ hospitals across the US.
With access to MIMIC, can access eICU-CRD immediately after signing an updated DUA.
paper
Other lists or pooling resources (relevant xkcd)
- Giorgos Sfikas: medical imaging datasets on github
- Andy Beam: medical data on github
- Christopher Madan: openMorph (open-access MRI, well structured list)
- Stephen Aylward's list of open-Access Medial Image Repositories
- google dataset search
- grand-challenges
- academic torrents
- multiBrain
- openneuro databse
Note the nice "fast preview" feature - The Cancer Image Archive
- Cornell Public Image Databases