SST fMRI Brain Decoding

Semi Spatio-Temporal fMRI Brain Decoding

The ability to decode mental states has been referred to as “Mind Reading”, “Reading/Seeing the Dreams” by some journalists and mass interest in  this field in the society is evident. fMRI Brain decoding is accomplished via classification of a subjects mental states given fMRI signals acquired while the subject is in certain mental tasks. Despite the novelty of this field, its potential capabilities has drawn many researchers to work in this field.

In an fMRI experiment the subject is directed to do certain mental tasks being fed with different stimuli, while the MRI machine scans the subject’s brain. fMRI captures the mental activity at different voxels (volumetric pixels) via measuring the amount of oxygenation of hemoglobin. Most models used in this field are linear models since they yield interpretable results (classifier weights can be inferred as the spatial significance of the activated regions in brain) and the satisfactory results obtained using them (both on the computation side and the final classification accuracy). Two major priors used in brain decoding are spatial smoothness and sparsity in the classifier weights. These could be explained as : we believe for a mental activity a few neighboring parts of brain are activated. Even though mental tasks include temporal changes in fMRI signals, most studies in this field derive their model upon single snapshots (volumes) of brain. Our research includes providing a model that can utilize the current spatial techniques while using temporal information in fMRI signals.

Related Publication:

  • Mohammad Hadi Kefayati, Hamid Sheikhzadeh, Hamid R. Rabiee, Ali Soltani-Farani, “Semi-Spatiotemporal fMRI Brain Decoding”, 3rd International Workshop on Pattern Recognition in NeuroImaging, University of Pennsylvania, Philadelphia PA, USA, 2013 (pdfposter)

Project code:

  • This code is written for MATLAB and contains routines for semi-spatio temporal method used in the above publication. Please cite the above work if you use this software and contact first author in case of any problems.

People involved:

  • Mohammad Hadi Kefayati, Ali Soltani-Farani, Hamid R. Rabiee

Sparse Signal Processing Group