Learning Deep Temporal Representations for fMRI Brain Decoding

2015-07-11
Firat, Orhan
Aksan, Emre
Oztekin, Ilke
Yarman Vural, Fatoş Tunay
Functional magnetic resonance imaging (fMRI) produces low number of samples in high dimensional vector spaces which is hardly adequate for brain decoding tasks. In this study, we propose a combination of autoencoding and temporal convolutional neural network architecture which aims to reduce the feature dimensionality along with improved classification performance. The proposed network learns temporal representations of voxel intensities at each layer of the network by leveraging unlabeled fMRI data with regularized autoencoders. Learned temporal representations capture the temporal regularities of the fMRI data and are observed to be an expressive bank of activation patterns. Then a temporal convolutional neural network with spatial pooling layers reduces the dimensionality of the learned representations. By employing the proposed method, raw input fMRI data is mapped to a low-dimensional feature space where the final classification is conducted. In addition, a simple decorrelated representation approach is proposed for tuning the model hyper-parameters. The proposed method is tested on a ten class recognition memory experiment with nine subjects. Results support the efficiency and potential of the proposed model, compared to the baseline multi-voxel pattern analysis techniques.

Suggestions

Statistical disease detection with resting state functional magnetic resonance imaging
Öztürk, Ebru; İlk Dağ, Özlem; Department of Statistics (2017)
Most of the functional magnetic resonance imaging (fMRI) data are based on a particular task. The fMRI data are obtained while the subject performs a task. Yet, it's obvious that the brain is active even when the subject is not performing a task. Resting state fMRI (R-fMRI) is a comparatively new and popular technique for assessing regional interactions when a subject is not performing a task. This study focuses on classifying subjects as healthy or diseased with the diagnosis of schizophrenia by analyzing ...
ASSESSMENT OF RPIM SHAPE PARAMETERS FOR SOLUTION ACCURACY OF 2D GEOMETRICALLY NONLINEAR PROBLEMS
BOZKURT, ÖMER YAVUZ; KANBER, BAHATTİN; Aşık, Mehmet Zülfü (World Scientific Pub Co Pte Lt, 2013-06-01)
This study discussed the effects of shape parameters on the radial point interpolation method (RPIM) accuracy in 2D geometrically nonlinear problems. Four finite deformation problems with compressible Neo-Hookean material are numerically solved with the RPIM algorithm using the multi-quadric (MQ) radial basis function. Both regular and irregular node distributions are used. Their displacements and Cauchy stresses are compared for different values of shape parameters and monomial basis. It is found that the ...
Classification in Frequency Domain of EEG Signals of Motor Imagery for Brain Computer Interfaces
Halıcı, Uğur (2009-05-22)
In this study the classification of the EEG signals recorded during motor imagery for curser movement in brain computer interfaces is examined, in which the feature vectors obtained in frequency domain is used and then the linear transformations are applied for reducing the size of the feature vectors.
Neural networks with piecewise constant argument and impact activation
Yılmaz, Enes; Akhmet, Marat; Department of Scientific Computing (2011)
This dissertation addresses the new models in mathematical neuroscience: artificial neural networks, which have many similarities with the structure of human brain and the functions of cells by electronic circuits. The networks have been investigated due to their extensive applications in classification of patterns, associative memories, image processing, artificial intelligence, signal processing and optimization problems. These applications depend crucially on the dynamical behaviors of the networks. In t...
Time reconstruction and performance of the CMS electromagnetic calorimeter
Chatrchyan, S.; et. al. (IOP Publishing, 2010-03-01)
The resolution and the linearity of time measurements made with the CMS electromagnetic calorimeter are studied with samples of data from test beam electrons, cosmic rays, and beam-produced muons. The resulting time resolution measured by lead tungstate crystals is better than 100 ps for energy deposits larger than 10 GeV. Crystal-to-crystal synchronization with a precision of 500 ps is performed using muons produced with the first LHC beams in 2008.
Citation Formats
O. Firat, E. Aksan, I. Oztekin, and F. T. Yarman Vural, “Learning Deep Temporal Representations for fMRI Brain Decoding,” 2015, vol. 9487, Accessed: 00, 2020. [Online]. Available: https://hdl.handle.net/11511/38910.