All Challenges
Here is an overview of all challenges that have been organized within
the area of medical image analysis that we are aware of. If you know
any study that would fit in this overview, or want to advertise your
challenge, please send an email to support@grand-challenge.org and we
will add the challenge to the list on this page.
Segmentation of THoracic Organs at Risk in CT images
SPIE-AAPM-NCI BreastPathQ:Cancer Circularity Challenge: Participants will be tasked to develop an automated method for analyzing histology patches extracted from whole slide images and assign a score reflecting cancer cellularity for tumor burden assessment in each.
Automated measurement of fetal head circumference using 2D ultrasound images
Classification of clinical significance of prostate lesions using multi-parametric MRI data
Peripheral Artery:Vein Enhanced Segmentation (PAVES) is the challenge focussed on providing easily interpretable and relevant images that can be readily understood by clinicians (vascular interventional radiologists & vascular surgeons) from MRA datasets where the venous and arterial vasculature may be equally enhanced. The setting is lower limb arterial occlusive disease where imaging of the below knee arterial vasculature is critical in planning limb salvage interventions. However, the competing demands of the high spatial resolution needed to image small vessels versus imaging time constraints where there is often a very short arteriovenous transit time for contrast passage form arterial to venous compartments makes imaging challenging. While dynamic MRA techniques can usually allow arterial imaging without venous ‘contamination’ these necessarily sacrifice spatial resolution.
MICCAI 2018 challenge for Multi-organ nuclei segmentation from H&E stained histopathological images.
The MSD challenge tests the generalisability of machine learning algorithms when applied to 10 different semantic segmentation task.
The goal of the Retinal Fundus Glaucoma Challenge (REFUGE) is to evaluate and compare automated algorithms for glaucoma detection and optic disc/cup segmentation on a common dataset of retinal fundus images.
MICCAI Challenge 2018 for Correction of Brainshift with Intra-Operative Ultrasound
Submit your method for lung and brain registration on https://github.com/SuperElastix/SuperElastix! Your method is easily accessible to end-users and automatically compiled, tested, and benchmarked weekly on several different data sets.
The LUMIC challenge tests the accuracy in registration between pre- and post-contrast CT chest images for algorithms, using an anthropomophic digital phantom.
This challenge evaluates automated techniques for analysis of fundus photographs. We target segmentation of retinal lesions like exudates, microaneurysms, and hemorrhages and detection of the optic disc and fovea. Also, we seek grading of fundus images according to the severity level of DR and DME.
Can you develop a method for automatic detection of cancerous regions in breast cancer histology images?
As an endoscopic vision CAI challenge at MICCAI, our aim is to provide a formal framework for evaluating the current state of the art, gather researchers in the field and provide high quality data with protocols for validating endoscopic vision algorithms.
MURA (musculoskeletal radiographs) is a large dataset of bone X-rays. Algorithms are tasked with determining whether an X-ray study is normal or abnormal.
Automatically detect a visual signal for pneumonia in chest x-rays.
Assess the probability of a TB patient having resistant form of tuberculosis based on the analysis of chest CT scan, automatically categorize each TB case into one of the following five types: (1) Infiltrative, (2) Focal, (3) Tuberculoma, (4) Miliary, (5) Fibro-cavernous and assessing TB severity score.
The aim of the challenge is the systematic evaluation of the performance of algorithms that enable the harmonisation of muti-shell DW MRI data from clinically available Siemens Prisma Scanner and the state-of-the-art Siemens Connectome Scanner. On the one hand, algorithms will be tested for their ability of harmonising the same DW MRI protocol across different MRI scanners. On the other hand, algorithms will be tested for their ability of enhancing the quality of routine DW MRI data sets to resemble the latest state-of-the-art acquisitions, via extrapolation of information that was not acquired given a set of training examples.
The goal of this recurring challenge is to help participants develop image analysis tools to enable the automated diagnosis of melanoma from dermoscopic images. There are three separate tasks: lesion segmentation, lesion attribute detection and disease classification.
The goal of MICCAI 2018 Challenge on Automatic IVD Localization and Segmentation from 3D Multi-modality MR (M3) Images is to investigate (semi-)automatic IVD localization and segmentation algorithms and provide a standard evaluation framework with a set of multi-modality MR images acquired with Dixon protocol.
The purpose of this challenge is to directly compare methods for segmentation of gray matter, white matter, cerebrospinal fluid, and other structures on 3T MRI scans of the brain, and to assess the effect of (large) pathologies on segmentation and volumetry.
Contestants are tasked to predict, using primary tumor 18F-FDG PET-derived radiomics features +/- matched clinical data, whether a tumor arising from the oropharynx will be controlled by definitive radiation treatment (RT).
Create an algorithm to automate nucleus detection from a variety of brightfield and fluorescence images.
This challenge addresses the validation of 3D diffusion MRI fiber tractography. The task is to provide the most accurate reconstruction of fiber pathways in both a physical phantom and real brain tissue.
Classify a cohort of lower grade glioma tumor cases into two sub-types: Oligodendroglioma and Astrocytoma when information from two types of imaging data – Radiology images and Pathology images – is used.
BraTS 2018 utilizes multi-institutional pre-operative MRI scans and focuses on the segmentation of intrinsically heterogeneous (in appearance, shape, and histology) brain tumors, namely gliomas. Furthemore, to pinpoint the clinical relevance of this segmentation task, BraTS’18 also focuses on the prediction of patient overall survival, via integrative analyses of radiomic features and machine learning algorithms.
ISLES 2018 asks for methods that allow the segmentation of stroke lesions based on acute CT perfusion data. A data set of 103 stroke patients and matching expert segmentations are provided.
Detect and segment all the nuclear material in a given set of image tiles extracted from whole slide tissue images.
This challenge is organised by the French Society on Radiology. After a first phase of collecting medical imaging across France, different challenges will be launch on different organs and different modalities. The results of the challenge will be announced during the French Days of Radiology (12-15 October).
Predict overall survival based on predictors derived from contrast-enhanced liver CT scans and patient clinical variables.
Develop a generalizable learning algorithms for 10 different semantic segmentation tasks.
Retinal OCT Classification Challenge (ROCC) is organized as a one day Challenge in conjunction with MVIP2017. The goal of this challenge is to call different automated algorithms that are able to detect DR disease from normal retina on a common dataset of OCT volumes, acquired with Topcon SD-OCT devices.
This challenge will provide a platform for comparison of various automatic/semi-automatic segmentation algorithms when they are used to delineate organs at risk from CT images for thoracic radiation treatment planning.
The Alzheimer’s Disease Prediction Of Longitudinal Evolution (TADPOLE) challenge is brought to you by the EuroPOND consortium in collaboration with the Alzheimer’s Disease Neuroimaging Initiative (ADNI).
The Challenge on Automatic Tool Annotation for cataRACT Surgery aims at evaluating image-based tool detection algorithms in the context of the most common surgical procedure in the world.
The goal of the Retinal OCT Fluid Challenge (RETOUCH) is to compare automated algorithms that are able to detect and segment various types of retinal fluid lesions on a common dataset of optical coherence tomography (OCT) volumes representing different retinal diseases, acquired with devices from different manufacturers.
Automated detection and classification of breast cancer metastases in whole-slide images of histological lymph node sections. This task has high clinical relevance and would normally require extensive microscopic assessment by pathologists.
Assess the probability of a TB patient having resistant form of tuberculosis based on the analysis of chest CT scan, and automatically categorize each TB case into one of the following five types: Infiltrative, Focal, Tuberculoma, Miliary, Fibro-cavernous
Segment, analyze and diagnose skin cancer from dermoscopic images provided by the International Skin Imaging Collaboration.
Can you improve the detection of lung cancer? Given a CT scan, predict if someone has lung cancer. The 2017 Data Science Bowl has $1 million in prizes.
Offline challenge for the diagnostic classification of clinically significant prostate lesions using quantitatve image analysis methods. Continued live in https://prostatex.grand-challenge.org/
ISLES 2017 asks for methods that predict stroke lesion outcome based on acute MRI data. A multi-spectral data set of 48 stroke patients and matching expert segmentations are provided.
BraTS 2017 utilizes multi‐institutional pre‐operative MRI scans and focuses on the segmentation of intrinsically heterogeneous (in appearance, shape, and histology) brain tumors, namely gliomas. Furthermore, in order to pinpoint the clinical relevance of this segmentation task, BraTS’17 also focuses on the prediction of patient overall survival, via integrative analyses of radiomic features and machine learning algorithms.
Develop an algorithm which accurately identifies a woman’s cervix type from images. Doing so will prevent ineffectual treatments and helps referral for cases that require more advanced treatment.
This challenge directly compares methods for the automatic segmentation of white matter hyperintensities of presumed vascular origin from FLAIR and T1-weighted MR images.
Tissue microarrays (TMAs) can provide new biomarkers that could be of value in diagnosis, predicting outcome and response to therapy. Goal of this challenge is to build predictions models for thyroid cancer from TMAs.
This challenge investigates progress in cerebellum lobule segmentation and labeling, structured around three MRI datasets.
iSeg-2017 compares (semi-)automatic algorithms for the segmentation of 6-month infant brain tissues and the measurement of corresponding structures using T1- and T2-weighted brain MRI scans.
Offline challenge to predict the Gleason Grade for a prostate MRI exam consisting of axial and sagittal T2-weighted images, Ktrans images (computed from dynamic contrast-enhanced), and apparent diffusion coefficient (ADC) images. Continued in https://prostatex.grand-challenge.org/
The goal of TraCED is to assess the reproducibility of common and emerging fiber tracking pipelines/algorithms using clinically feasible MR imaging sequences.
CoronARe compares state-of-the-art methods in terms of 3D reconstruction accuracy of coronary arteries from interventional C-arm rotational angiography.
Develop an algorithm which can most accurately determine skeletal age on a validation set of pediatric hand radiographs, using a large data set of over 12,000 images.
Develop an algorithm for multi-modality whole heart segmentation from cardiac CT and MRI images.
We encourage researchers to develop automatic segmentation algorithms to segment liver lesions in contrast-enhanced abdomen CT scans. The data and segmentations are provided by various clinical sites around the world.
Develop an algorithm for MRI cardiac segmentation AND/OR predict pathology of different patients given cardiac MRI images.
The LUNA16 challenge will focus on a large-scale evaluation of automatic nodule detection algorithms for chest CT.
ISBI Endoscopic Image Challenge
The goal of this challenge is to evaluate new and existing algorithms for automated detection of cancer metastasis in digitized lymph node tissue sections. Two large datasets from both the Radboud University Medical Center and the University Medical Center Utrecht are provided.
Quantitatively assess the diagnostic performance of denoising and iterative reconstruction techniques on common low-dose patient CT datasets using a detection task.
The ISLES 2016 challenge treats the challenging problem of stroke lesion/clinical outcome prediction from acute MRI scans. They provide acute MRI scans of a large number of acute stroke cases and associated clinical parameters. The associated ground truth is the final lesion volume (Task I) as manually segmented in 3 to 9 month follow-up scans, and the clinical mRM score (Task II) denoting the degree of disability.
MS lesions segmentation as part of the effort towards automatic segmentation of MRI scans of MS patients.
Focus on two key unaddressed areas in the field of localisation microscopy: 3D and Ultra High-density 2D intended for live cell applications.
This workshop brings together two MICCAI-Workshop series (CVII and STENT) in order to form a common event on technological and scientific research concerned with endovascular procedures.
The overarching goal of the challenge is to develop image analysis tools to enable the automated diagnosis of melanoma from dermoscopic images.
The increase in volume and variety of multi modal cancer data along with advances in computational power and analytics have given rise to challenges and opportunities in determining optimal strategies for development of systems for clinical decision support and advancement of precision medicine.
To develop a surgical workflow monitoring system by detecting both the surgical workflow and identifying all surgical tools that are present in an image.
This challenge focuses on algorithms for segmentation of left atrial wall thickness from MRI and CT.
BigNeuron is a community effort to define and advance state-of-the-art of single neuron reconstruction: an essential unsolved challenge in brain science.
Declining cardiac function is a key indicator of heart disease. The 2015 Data Science Bowl challenges you to create an algorithm to automatically measure end-systolic and end-diastolic volumes in cardiac MRIs.
Automatic Intervertebral Disc Localization and Segmentation from 3D Multi-modality MR (M3) Images.
Evaluate methods that predict the tumor proliferation score directly from whole-slide images.
This challenge is about segmenting the blood pool and myocardium from a 3D cardiovascular magnetic resonance images.
Its aim is to create a common ground to compare methods to find predictive MRI features, which help to characterise and distinguish mild traumatic brain injury patients from each other and healthy subjects.
Identify nerve structures in ultrasound images of the neck. The task in this competition is to segment a collection of nerves called the Brachial Plexus (BP) in ultrasound images.
The tasks of the xVertSeg challenge are: To segment vertebrae from the given spine images that include fractured and non-fractured cases. To classify vertebrae from the given spine images into fractured and non-fractured cases along with specific morphological grades and cases of vertebral fractures.
The goal of this challenge is to evaluate algorithms for automatic reconstruction of neurons and neuronal connectivity from serial section electron microscopy data.
Automated segmentation of PET images for the delineation of tumor volumes
ImageCLEFmed 2016 consist of two tasks: analyze compound figures of the biomedical literature and provide a hierarchical classification of body parts in bone x-ray images.
Develop algorithms to classify screening digital mammography images as low and high risk for developing cancer. Thousands of digitized mammograms and outcome data are available for training.
Multimodal Brain Tumor Segmentation (BraTS), making available a large dataset of brain tumor MR scans in which the relevant tumor structures have been delineated.
Investigates cell detection and cell segmentation for the automated analysis of multi-layer cytology volumes.
Gland morphology is used to assess the degree of malignancy of several adenocarcinomas, including prostate, breast, lung, and colon. This challenge evaluates gland segmentation algorithms on images of H&E stained slides, consisting of a variety of histologic grades.
A task targeted at the identification, multi-label classifcation and separation of compound figures from biomedical literature.
Compare and evaluate state-of-the-art whole-cell and nucleus tracking methods using both real (2D and 3D) time-lapse microscopy videos of labeled cells and nuclei, along with computer generated video sequences simulating nuclei moving in realistic environments
Evaluates and investigates automated segmentation of gray matter, white matter and cerebrospinal fluid on both the NeoBrainS12 and MRBrainS13 datasets.
Investigates automated methods for detection of caries in 120 bitewing X-ray images collected from 120 patients.
Evaluates quantitative image-analysis methods for the diagnostic classification of malignant and benign lung nodules. Training set to be released November 21st.
The VISCERAL Anatomy3 challenge is a continuation and extension of two earlier challenges on whole-body labelling in 3D medical images. Its goal is the segmentation of abdominal organs and the localization of anatomical landmarks.
Benchmarks automated detection and statistical analysis on cephalometric x-ray images of 400 patients.
Competing teams will apply their automatic lesion segmentation algorithms to MR neuroimaging data acquired at multiple time points from MS patients. Algorithms will be evaluated against manual segmentations from multiple raters in terms of their segmentation accuracy and ability to track lesion evolution.
Requires only a written proposal describing specific and sensitive biomarker(s) that are highly associated with liver fibrosis and could be used as a surrogate for clinical efficacy and ideally, could guide treatment selection.
Models the statistical shape of the left ventricle (LV) to detect myocardial infarction. training comprises 100 cases with myocardial infarction 100 asymptomatic cases from the Cardiac Atlas Project.
Identify signs of diabetic retinopathy in eye images. This Kaggle competition offers 100,000 dollars in prize awards.
Presents the current state-of-the-art in automated tracking of anatomical landmarks in the liver and compares different methods
Brings together and furthers research in the area of retinal cyst segmentation, a key indicator of sight degrading diseases such as RVO and AMD, in SD-OCT ophthalmic imaging. Organised by the OPTIMA lab, Medical University of Vienna.
Benchmarks automatic segmentation performance. Provides 25 CT images with manually delineated data of brainstem, mandible, left and right optic nerves, optic chiasm, left and right parotid glands and submandibular glands.
Aims to identify the mathematical model for diffusion MRI that best describes the signal from in-vivo Human Brain White Matter acquired on the Connectom scanner with gradients of up to 300mT/m.
ISLES 2015 is a two-task challenge of automatic methods for sub-acute ischemic stroke lesion segmentation and acute stroke penumbra estimation. Versatile multi-center data, two expert segmentations and extensive automatized evaluation.
The task is to complete 10 partial liver segmentation from a set of segmented training datasets.
Multimodal Brain Tumor Segmentation (BraTS), making available a large dataset of brain tumor MR scans in which the relevant tumor structures have been delineated. Each data set has T1 MRI, T1 contrast-enhanced MRI, T2 MRI, and T2 FLAIR MRI volumes
Evaluates multi-atlas segmentation algorithms on a newly created and released standardized dataset of abdominal and pelvic anatomy on clinically acquired CT
This challenge evaluates and compares tractography pipelines: provide the best possible reconstruction of fiber pathways in a realistically simulated whole brain diffusion-weighted MR image.
The first challenge concerns vertebral fracture analysis, the second challenge is on intervertebral disc localization and segmentation.
MITOS & ATYPIA 14 Contest, hosted by conference ICPR 2014
Detection of mitosis and evaluation of nuclear atypia on breast cancer H&E stained images
We seek algorithms that perform multi-class classification of patients with Alzheimer’s disease (AD), patients with mild cognitive impairment (MCI) and healthy controls (CN) using multi-center structural MRI data.
This workshop evaluates the performances of tractography algorithms in the reconstruction of peritumoral anatomy and corticospinal tract trajectory on pre-operative and post-operative diffusion data from patients with tumors in or near the motor system.
Provides a database of 10 spine CT scans along with manual segmentation of the thoracic and lumbar vertebrae for development, training and testing of spine segmentation algorithms. Five more data sets will be provided for the evaluation phase of the challenge.
Multimodal Brain Tumor Segmentation (BraTS). It contains three sub-challenges. 1: Segmentation of brain tumor (GBM) and its components, 2: Longitudinal Evaluation of time series image data and 3: Classification into Low- and and High Grade (GBM) gliomas.
Detection and segmentation of overlapping cells is a major hurdle for robust cervical cell analysis. In this challenge, the targets are to extract the boundaries of individual cytoplasm and nucleus from Pap smear microscopy images.
This challenge compares and evaluate state-of-the-art whole-cell and nucleus tracking methods using both real and artificial 2D and 3D time-lapse microscopy videos of fluorescently labeled cells and nuclei moving in realistic environments.
This challenge presents the current state-of-the-art in automated tracking of anatomical landmarks in liver ultrasound images and compare different methods.
The orCaScore challenge compares methods for automatic and semi-automatic coronary artery calcium scoring in cardiac CT scans.
In this benchmark we will evaluate segmentation and detection algorithms on a large dataset of clinical wide-field-of-view MRI and CT scans. Major organs and substructures have been manually delineated, and anatomical interest points or landmarks have been set.
Deconvolution is one of the most common image-reconstruction tasks in 3D fluorescence microscopy. This challenge benchmarks existing deconvolution algorithms and to stimulate the community to look for novel, global and practical approaches to this problem.
The Liver CT Annotation Task targets automatic annotation of 3D liver CT data by filling a form of imaging observations at the liver, vessel and lesion levels. The form is built based on a novel open source ontology (ONLIRA) of liver in radiology.
The Shape Symposium hosts the first Grand Challenge in Statistical Shape Modeling. The task is to build a model of the liver from a set of segmented training datasets.
The goal of this challenge is to investigate a suitable automatic landmark detection technology for cephalometric x-ray images and provide a standard evaluation framework with a clinical dataset.
The evaluation of osteoporotic disease from bone radiograph images presents a major challenge for pattern recognition and medical applications. The goal is to identify osteoporotic cases from healthy controls on 2D bone radiograph images, using texture analysis.
Machine-learning practitioners can try out their favorite (novel or not) supervised learning algorithm(s) on brain imaging (neuroimaging) data for clinically relevant prediction problems.
Evaluation of classification of LGG and GBM, and segmentation of necrotic and normal brain regions from high-res digital pathology slide clinical cases.
This challenge compares left ventricle segmentation methods for both End Diastolic and End Systolic phase instances in 3D cardiac ultrasound.
The objective is to detect various lesions in anatomical regions such as the bones, liver, brain, lung, or lymph nodes in CT and MR volumes. There are overall 308 annotated lesions annotated through medical experts.
We are presenting results of our segmentation method for subsolid lung nodules.
CRASS stands for Chest Radiograph Anatomical Structure Segmentation. The challenge currently invites participants to send in results for clavicle segmentation algorithms.
The aim of this challenge is to benchmark existing deconvolution algorithms and to stimulate the community to look for novel, global and practical approaches to this problem.
Mitotic activity is one of the strongest prognosticators for invasive breast carcinoma. This challenge evaluates and compares (semi-)automatic mitotic figure detection methods on regions extracted from whole-slide pathology images.
VISCERAL is an EU funded project in which various benchmarks are organized. The first Benchmark focuses on whole body labelling in 3D medical imaging data.
Automated Segmentation of Prostate Structures (ASPS)
The MRBrainS evaluation framework compares algorithms for segmentation of grey matter, white matter and cerebrospinal fluid on multi-sequence (T1-weighted, T1-weighted inversion recovery and FLAIR) 3 Tesla MRI scans of the brain.
Multimodal Brain Tumor Segmentation (BraTS), making available a large dataset of brain tumor MRscans in which the tumor and edema regions have been manually delineated,adding another 20 multimodal image volume from high and low grade gliomapatients to the BRATS 2012 data set. All images – in both the publiclydistributed training data set, and the blinded test data set- are annotatedthrough clinical experts who annotated four different types of tumor substructurs(edema, enhancing core, non-enhancing core, necrotic core).
Manually tracking cells is an extremely laborious task, due to the large amount of image data acquired during live-cell studies. This challenge compares automatic approaches for cell tracking.
The goal of this challenge is to compare interactive and (semi)-automatic segmentation algorithms for MRI of the prostate.
The VESSEL12 challenge compares methods for automatic (and semi-automatic) segmentation of blood vessels in the lungs from CT images.
High angular resolution diffusion imaging 2012
This is a segmentation challenge for automatically segmenting anatomical structures to measure standard obstetric biometric parameters, from 2D fetal ultrasound images.
Multimodal Brain Tumor Segmentation (BraTS), making available a large dataset of brain tumor MR scans in which the tumor and edema regions have been manually delineated. In addition, we also provide realistically generated synthetic brain tumor datasets for which the ground truth segmentation is known.
Quantitative analysis of dynamic processes in biological cells requires accurate tracking of large numbers of particles in time-lapse microscopy images. This challenge objectively compares the performance of particle tracking algorithms for this purpose.
In this challenge, a full stack of EM slices will be used to train machine learning algorithms for the purpose of automatic segmentation of neural structures.
The aim of the NeoBrainS12 challenge is to compare (semi-)automatic algorithms for segmentation of neonatal brain tissues (cortical and central grey matter, non-myelinated and myelinated white matter, brainstem and cerebellum, and cerebrospinal fluid in the ventricles and in the extracerebral space) using T1- and T2-weighted MRI scans of the brain.
The goal of LOLA11 (LObe and Lung Analysis 2011) is to compare methods for (semi-)automatic segmentation of the lungs and lobes from chest computed tomography scans. Any team, whether from academia or industry, can join.
The objective of this challenge is to evaluate the accuracy and reproducibility of different motion analysis algorithms applied to a multimodal cardiac database. The database includes a dataset from a dynamic phantom1 and 15 datasets from healthy volunteers.
This challenge provides IVUS pullback datasets from different frequencies with various degrees of difficulty. Aim is to evaluate segmentation of the inner wall (lumen) and outer wall (media/adventitia).
Knee cartilage segmentation is a clinically relevant segmentation problem which has gained considerable importance in recent years. The goal of SKI10 is to compare different algorithms for cartilage and bone segmentation from knee MRI data.
The goals of this competition were to raise awareness of the problem of automated neuronal reconstruction, spur development of automated and semi-automated algorithms, and to gauge the state of the art in the field.
Accurate registration of thoracic CT is both extremely useful in clinical terms and exceptionally challenging due to the elastic nature of lung tissue deformations. EMPIRE10 provides a platform for in-depth evaluation and fair comparison of available registration algorithms for this application.
ANODE09 is an initiative to compare systems that perform automatic detection of pulmonary nodules in chest CT scans on a single common database, with a single evaluation protocol.
The goal of the EXACT study is to compare algorithms to extract the airway tree from thoracic computed tomography scans using a common dataset and performance evaluation method.
VOLCANO'09: Volume Change Analysis of Nodules
2009 Prostate Segmentation Challenge
ROC: Retinopathy Online Challenge
The goal of CAUSE07 is to compare different algorithms to segment the caudate nucleaus from brain MRI scans.
The goal of this competition is to compare different algorithms to segment the liver from clinical 3D computed tomography (CT) scans.
