A Machine learning neuroimaging challenge for automated diagnosis of MCI

I have written in this blog about the importance of having shared datasets on which algorithms could be validated (link).
This is fundamental to obtain comparable performances, which is particularly important when we are dealing with medical systems, such as those used to automatically diagnose neurodegenerative diseases.

In these months, “A Machine learning neuroimaging challenge for automated diagnosis of Mild Cognitive Impairment (MCI)” is being held on the Kaggle platform.
This challenge makes use of data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI), asking competitors to automatically classify patients belonging to four different classes: Alzheimer’s Disease (AD), MCI-converter to AD (cMCI), MCI-not converter to AD (ncMCI), and Healthy Controls (HC).

So, a four-label classification problem.

Considering that the accuracy for random guessing is 25%, we are curious to see what will be the accuracy reached by the best team.

So, stay tuned! [to be updated]