About

WEKA is a Java-based software platform designed for the implementation, distribution and application of classifiers and methods for regression analyses for all kinds of data. Unfortunately, technical issues and some pecularities of fMRI data prevented the import and classification of neuroimaging data in this software. Here, we present a tool for the integration of fMRI data in WEKA, and thereby provide a way to access a broad collection of recent analysis tools for data classification.

Download

Version: 18.11.2011
Nifti Converter (10.8 MB)
Sourcecode (9.7 MB)

Furthermore, you need Java3D installed on your system.

Instruction

The MRI-Converter assumes a certain structure of the data to be classified. All images (and only those images) of one class should be in one folder and all class folders should be in one parent-folder. The converter will generate the arff-file for WEKA in this parent folder. If the class variable specifies an intra-subject variable, the images of the corresponding subjects should have the same order in all class-directories.

You should increase the heap space of your system when you run the converter. For example, you could start the converter with

java -Xmx1400M -jar MRIConverter.jar

Enter the name of the WEKA-file (without the ending) that should be created

You can directly proceed with the conversion of the data. Click on the Convert-Button and enter the folder that contains the sub-directories with the fMRI-data

The converter offers some preprocessing steps to reduce the size of the data before conversion. These can help to avoid memory allocation and runtime problems.