HomeWhy Challenges?All ChallengesCreate your own projectContributors
Sign in / Register

Welcome to the web site of the challenge on Computer-Aided Diagnosis of Dementia based on structural MRI data.

Classification methods for dementia have demonstrated high performances, but are not yet used in clinical practice. A major reason for this is that these methods are not well enough validated for clinical use.

In this challenge, we aim to take a step forward to the clinical use of computer-aided diagnosis methods for dementia by performing a large-scale objective validation. To compare the performance of image-based diagnosis methods, all researchers are invited to participate with their algorithms.

How does it work?

We provide a test set of multicenter clinical-representative T1-weighted MRI data of patients with Alzheimer's disease (AD), mild cognitive impairment (MCI) and healthy controls. The diagnosis methods can be trained and tuned on any suitable data (i.e. from the ADNI database). Additionally, we provide a small set of training data with diagnostic labels. If you would like to participate in this challenge, you can register a team on this web site. To be able to download the data from this web site, you are requested to email a description of your algorithm and to sign the data usage agreement. Subsequently you can run your classification algorithm on the data and submit the obtained diagnostic labels. Each submission is evaluated against the reference standard (clinical diagnosis) and the results are published on this web site.

Motivation

This challenge is designed to address comparability, generalizability, and clinical applicability:

  • Standardized evaluation framework: the use of the same data set and evaluation methods enables better and more objective comparison of the diagnosis methods.
  • Previously unseen multicenter data: the ground truth diagnoses are blinded to the participants to avoid overtraining and to promote generalizability of the methods.
  • Multi-class classification of AD, MCI and controls: according to the current clinical standards, a multi-class diagnosis is required.
 
The organizers,
 
Esther E. Bron, Stefan Klein, Marion Smits, John C. van Swieten, Wiro J. Niessen
Erasmus MC, the Netherlands.

News

18 September 2014: The workshop was a success, congratulations to all participants. Results are online now.
 

2 September 2014: CADDementia is nominated for the Dutch Data Award!


1 March 2014: Data is available for download

• • • • • • • • • • • • • • • • • • • • • • •

Event sponsors

 

• • • • • • • • • • • • • • • • • • • • • • •

Related challenge

The MICCAI 2014 Machine Learning Challenge is a related event that focuses specifically on machine-learning algorithms and provides ready-to-use image-based features.

 


Consortium for Open Medical Image Computing © 2012-2014