NOTE: [Update May 14, 2014] We recently realized that a few of the training subjects in the continuous regression problem were duplicates. We therefore recommend the use of ContTrain_sbj_list2.csv which contains a third column -- any subject with a Yes in that column can be safely discarded from training/cross-validation as its inclusion might bias/skew your results, particularly your cross-validation estimates.
Please refer to the data description page for detailed information on the data.
We have two problems: binary and continuous.
In each problem, we have two datasets: training and testing. For the training data, the target variables (i.e., labels) are made available. For the testing data, only the image data are provided.
There are two types of image data: summary and mri. The summary data are feature vectors (of length 184 per sample) that are a collection of morphological measurements, such as volume/thickness of an anatomical region, automatically computed from the MRI data. The mri data are the original, de-identified MRI scans, in nifi format.
To download the summary data, click here.
Under the summary folder, there are two sub-folders, binary and cont (i.e., continuous), corresponding to the two types of prediction problems.
Under these, you will find the image-derived summary data (both for training -BinaryTrain_data.csv- and testing -BinaryTest_data.csv-) and labels for the training data (BinaryTrain_sbj_list.csv). These files are comma separated text files. Each row is a separate subject/sample.
To download the MRI scans, click here.
Under the mri folder, there are two sub-folders, binary and cont (i.e., continuous), corresponding to the two types of prediction problems.
Under these, there are the MRI files, called Train*.nii.gz and Test*nii.gz, referring to the training and testing data, respectively. BinaryTrain_sbj_list.csv, which contains the labels for the training data, can also be found in these directories.
For Challenge Instructions, go here.