Validation data sets

Name

Organization

Main objective of the data sets

Clinical context

Data sets (modalities and features)

Realism vs. control

Reference

Availability

Web site

Truthcube

MGH , Boston USA

Validation of tissue deformation analysis

None

CT (deformed and undeformed data)

Simulated data

Control + / realism --

Gold standard available

 

http://www.medicalsim.org/truthcube/

RIRE

Vanderbilt University + NLM USA

To validate and compare rigid registration methods

Patient brain images

19 MR T1, T2, and PD, CT, PET neurological data sets

Clinical data

Control - / realism ++

GS now available

Validation metrics: TRE on VOI

Public

http://www.insight-journal.org/rire/

BrainWebColin27

MNI Montreal Canada

Validation of image processing methods

One normal subject with no, mild, moderate or severe Multiple Sclerosis lesions added (4 different phantoms from the same anatomy).

Simulated MR T1 T2 and PD neurological data sets normal and pathological (with one of the four phantoms and different TE, TR, noise, RF values)

Simulated data

Control + / realism +

Gold standard available

Public

http://mouldy.bic.mni.mcgill.ca/cgi/bw/submit_request

BrainWeb b)

MNI Montreal Canada

Validation of image processing methods

20 normal subject phantoms with 20 simulated t1-weighted MRI available

Simulated T1 weighted MR images

Simulated data

Control + / realism +

Simulated MRs with 12 anatomical labels

Public

http://mouldy.bic.mni.mcgill.ca/brainweb/anatomic_normal_20.html

OASIS

Washington University Alzheimer’s Disease Research Center

Validation of image processing algorithms trying to detect differences between MRI of AD patients and normal controls

Patients with dementia and Alzheimer’s disease and normal controls

T1 weighted MR images with clinical scores of 416 subjects for the cross sectional study and of 150 subjects for the longitudinal study.

Clinical data

Control - / realism ++

Clinical data is available

Public

http://www.oasis-brains.org/

Visible Human Project

The National Library of Medicine USA

Mainly used for segmentation evaluation and realistic visualization

None

MR CT and high resolution colour data sets of the whole body

Normal subject data

Control - / realism +

Reference available through multimodal images

Upon request

http://www.nlm.nih.gov/research/visible/visible_human.html

The Zubal Phantom

Yale University USA

Suitable for many computer-based modelling and simulation calculations as well as for segmentation evaluation

Head

Segmented CT and MR head and torso images

Normal subject data

Control - / realism +

Manually labelled data sets - Gold standard for segmentation

Public

http://noodle.med.yale.edu/zubal/

JSRT Database

Japanese Society of Radiological Technology Japan

Suitable for image processing, image compression, evaluation of image display, computer-aided diagnosis

Chest with and without nodules

Chest XRay with manually identified nodules along with clinical information

Clinical data

Control - / realism ++

Manually labelled data sets - Gold standard for detection

Upon request

http://www.jsrt.or.jp/web_data/english03.php

The SCR Database

Image Science Institute NL

Segmentation evaluation

Chest

247 segmented chest radiographs images

Clinical data

Control - / realism ++

Manually labelled lung fields, heart and clavicles - Gold standard for segmentation

Upon request

http://www.isi.uu.nl/Research/Databases/SCR/

The DRIVE Database

Image Science Institute NL

Validation of segmentation of blood vessels in retinal images evaluation

Subjects without and with signs of mild early diabetic retinopathy.

40 photographs of retinal vessels

Clinical data

Control - / realism ++

Manually and automatic labelled retinal vessels - Gold standard for segmentation

Upon request

http://www.isi.uu.nl/Research/Databases/DRIVE/

The GS Database

Image Science Institute NL

2D-3D image registration validation

Spinal

One data set with MR, CT, 3DRX, and fluoroscopic images of 2 spinal segments

Clinical data

Control - / realism ++

Geometrical transformations between data sets - Gold standard for registration

Upon request

http://www.isi.uu.nl/Research/Databases/GS/

Segmentation of the Liver Competition 2007

Image Science Institute NL

Validation of liver segmentation in CT images

Liver: patients with

tumors, metastasis and cysts in dierent sizes

20 training scans and 10 testing CT images

Clinical data

Control - / realism ++

Manually labelled data sets - Gold standard for segmentation

Upon request

http://www.sliver07.org/

CAUdate SEgmentation 2007

Image Science Institute NL

Validation of segmentation of caudate in MR images

Healthy controls and subjects

with Schizoptypal Personality Disorder

18 + 15 Brain MR images

Clinical data

Control - / realism ++

Manually labelled data sets - Gold standard for segmentation

Upon request

http://www.cause07.org/

The AV Database

Visages/U746 INRIA/INSERM, F

Validation of stereoscopic reconstruction, augmented virtuality, and surface registration

Image guided neurosurgery

Stereoscopic images of a physical phantom with associated CT scan and calibration matrices

Physical phantom

Control + / realism -

Computable transformations between images CS and CT CS - Gold standard for registration

Public

http://idm.univ-rennes1.fr/theme1/AV_validation/

The Non-Rigid Image Registration Evaluation Project (NIREP)

University of Iowa, USA

Validation of non rigid image registration

Brain images of 16 normal subjects

T1 MR brain images

Clinical data

Control + / realism +

Manual segmentation of  32 grey matter structures

Upon request

http://www.nirep.org/

The STARE Project

Clemson University, USA

Validation of optical nerve identification in retina images

31 healthy retinas and 50 retinas with disease

81 colour images of the retina

Clinical data

Control - / realism ++

Manually defined location of the center point of the nerve in each image

Public

http://www.parl.clemson.edu/stare/nerve/

The Internet Brain Segmentation Repository IBSR

Massachusetts General Hospital, USA

Validation of segmentation of brain structures in MR images

Normal subjects and patients with tumor

T1 MR Images of one adult subject, one 5 year old child, 38 normal subjects, and 2 patients with tumor

Clinical data

Control - / realism ++

Manual segmentations of different brain structures

Upon request

http://www.cma.mgh.harvard.edu/ibsr/data.html

The POPI-model

Creatis-LRMN, France

Validation of non-rigid registration between thorax CT images.

One patient with lung tumor

4D CT Thorax images (10*3D CT volume) taken at different times during breathing period

Simulated data

Control + / realism +

44 manual identified landmarks and results from their own registration method

Public

http://www.creatis.insa-lyon.fr/rio/popi-model

LPBA40

Laboratory of Neuro Imaging, UCLA, USA

Validation of MRI segmentation algorithms

Normal subjects

40 cases with MRI coregistered and corresponding anatomic probabilistic maps

Clinical data

Control - / realism ++

56 anatomical structures manually segmented

Upon request

http://www.loni.ucla.edu/Atlases/LPBA40

Alzheimer's Disease Neuroimaging Initiative (ADNI)

Laboratory of Neuro Imaging, UCLA, USA

Validation of image processing algorithms trying to detect differences between MRI of AD and MCI patients and normal controls

Controls, Alzheimer's Disease, Mild Cognitive Impairment

895 cases with MRI and PET

Clinical data

Control - / realism ++

Clinical data is available

Upon request

https://ida.loni.ucla.edu/services/Menu/IdaData.jsp?project=

LID (Lung Image Database)

Cancer Imaging Program, NCI, USA

Validation of Computer Aided Detection of  Lung cancer in CT images

Lung cancer

Spiral CT lung

 

Clinical data

Control - / realism ++

Marked-up annotated lesions

Public

http://imaging.cancer.gov/programsandresources/InformationSystems/LIDC/

National Biomedical Imaging Archive

Cancer Imaging Program, NCI, USA

Validation of Computer Aided Detection and classification of lesions from images

Different body parts

A large collection of medical image from different imaging protocols on different body parts. It includes LID data base.

Clinical data

Control - / realism ++

A lot in one place

Public

https://imaging.nci.nih.gov/ncia/login.jsf

DDSM (Digital Database for Screening)

University of South Florida, USA

Validation of methods for screening mammography

Breast cancer

2500 studies including 2 mammograms for each breast and patient clinical information

Clinical data

Control - / realism ++

Manually identified suspicious regions

Public

http://marathon.csee.usf.edu/Mammography/Database.html

Mini-MIAS (Mammographic Image Analysis Society)

University of Essex, GB

Validation of methods for screening mammography

Breast cancer

52 digitized mammograms

Clinical data

Control - / realism ++

Manually identified regions

Public

http://peipa.essex.ac.uk/info/mias.html

NCC-CIR

National Cancer Center, JP

More for clinicians. Could be used for CAD systems.

Lung and gastro-intestinal cancer

53 cases with medical images and diagnostic information. Images include radiographs, CT, magnetic resonance imaging (MR), endoscopic images, ultrasonographs and pathological images.

Clinical data

Control - / realism ++

Clinical data and diagnostic are available

Public

http://cir.ncc.go.jp/en/

MIDAS

Kitware/UNC

More for verification rather than validation

Normal and pathological subjects

A large collection of medical image from different imaging protocols on different body parts. It includes data from others databases (e.g., RIRE)

 

A lot in one place

Public

http://www.insight-journal.org/midas/

Pediatric MRI Data Repository

Multiple institutions, USA, CA

For studying brain development

Paediatric and young subjects

2 sub projects : 1) 450 from 5 to 18 years-old children, 140 from 0 to 5 year-old children ; longitudinal data with MR images, DTI and MRS

 

No real reference available.

Upon request

http://www.bic.mni.mcgill.ca/nihpd/info/

BIRN (Biomedical Informatics Research Network)

Multiple institutions, USA

For studying inter site variations. More a place to share data.

Healthy and pathological human brain, Cell, and Mouse atlases

A large collection of different DTI and fMRI brain images, Microscopy and MR Images based mouse atlases

 

No real reference available.

Public

http://www.nbirn.net/

International Consortium for Brain Mapping (ICBM)

Laboratory of Neuro Imaging, UCLA, USA

More a place to share data.

Brain, Normal Controls

851 cases with MRI, fMRI, MRA, DTI, PET

 

No real reference available.

Upon request

https://ida.loni.ucla.edu/services/Menu/IdaData.jsp?project=


Mono-subject MRI brain template