A Robust Normalization Method for fMRI Data for Brain Decoding

Date

2016-05-19

Author

Yildiz, Ozan
Dogan, Fethiye Irmak
GİLLAM, İLKE
Mizrak, Eda
Yarman Vural, Fatoş Tunay

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

29
views

0
downloads

Functional Magnetic Resonance Imaging (fMRI) methods produce high dimensional representation of cognitive processes under heavy noise due to the limitations of hardware and measurement techniques. In order to reduce the noise and extract useful information from the fMRI data, a sequence of pre-processing techniques, such as smoothing with spatial filters and z-scoring, are used. In this study, we suggest an additional normalization technique based upon a statistical property of fMRI data. We, first, define a random variable V(t) as the average voxel intensity value of a brain volume measured at a time instant t. Then, we measure the Pearson correlation between V(t) and 1/V(t) for all time instances. We observe that the Pearson correlation values are very close to -1, indicating that V(t) and 1/V(t) have a strong negative correlation. We show that one explanation for this property is V(t) being almost surely constant and the small fluctuations on V(t) caused by noise. The proposed method removes these fluctuations on the data resulting in almost surely constant brain volumes V(t) for all values of t. The effectiveness of the proposed normalization method is tested with well-known brain decoding algorithms and voxel selection methods. It is observed that the suggested normalization method improves the performance 1-2 percent on the average. The method also improves the signal to noise ratio.

Subject Keywords

Human Visual-Cortex, Representations, Patterns, Objects

URI

https://hdl.handle.net/11511/55707

DOI

https://doi.org/10.1109/SIU.2016.7496228

Conference Name

24th Signal Processing and Communication Application Conference (SIU)

Collections

Department of Computer Engineering, Conference / Seminar

Suggestions

OpenMETU
Core

A Sparse Temporal Mesh Model for Brain Decoding Afrasiyabi, Arman; Onal, Itir; Yarman Vural, Fatoş Tunay (2016-08-23) One of the major drawbacks of brain decoding from the functional magnetic resonance images (fMRI) is the very high dimension of feature space which consists of thousands of voxels in sequence of brain volumes, recorded during a cognitive stimulus. In this study, we propose a new architecture, called Sparse Temporal Mesh Model (STMM), which reduces the dimension of the feature space by combining the voxel selection methods with the mesh learning method. We, first, select the "most discriminative" voxels usin...
A Parametric Estimation Approach to Instantaneous Spectral Imaging Öktem, Sevinç Figen; Davila, Joseph M (2014-12-01) Spectral imaging, the simultaneous imaging and spectroscopy of a radiating scene, is a fundamental diagnostic technique in the physical sciences with widespread application. Due to the intrinsic limitation of two-dimensional (2D) detectors in capturing inherently three-dimensional (3D) data, spectral imaging techniques conventionally rely on a spatial or spectral scanning process, which renders them unsuitable for dynamic scenes. In this paper, we present a nonscanning (instantaneous) spectral imaging techn...
A new boundary element method formulation for the forward problem solution of electro-magnetic source imaging Tanzer, IO; Gençer, Nevzat Güneri (1997-11-02) Numerical solution of the potential and magnetic fields far a given electrical source distribution in the human brain is the essential part of electro-magnetic source imaging. In this study, the performance of Boundary Element Method (BEM) with different surface element types is explored. A new BEM formulation is derived that makes use of isoparametric linear and quadratic elements. It is shown that, quadratic elements provides superior performance over linear elements in terms of computation time and accur...
A Design for Real-time Neural Modeling on the GPU Incorporating Dendritic Computation Garaas, Tyler W.; Marino, Frank; Duzcu, Halil; Pomplun, Marc (2009-07-05) Recent advances in neuroscience have underscored the role of single neurons in information processing. Much of this work has focused on the role of neurons' dendrites to perform complex local computations that form the basis for the global computation of the neuron. Generally, artificial neural networks that are capable of real-time simulation do not take into account the principles underlying single-neuron processing. In this paper we propose a design for a neural model executed on the graphics processing ...
A temporal neural network model for constructing connectionist expert system knowledge bases Alpaslan, Ferda Nur (Elsevier BV, 1996-04-01) This paper introduces a temporal feedforward neural network model that can be applied to a number of neural network application areas, including connectionist expert systems. The neural network model has a multi-layer structure, i.e. the number of layers is not limited. Also, the model has the flexibility of defining output nodes in any layer. This is especially important for connectionist expert system applications.

Citation Formats

O. Yildiz, F. I. Dogan, İ. GİLLAM, E. Mizrak, and F. T. Yarman Vural, “A Robust Normalization Method for fMRI Data for Brain Decoding,” presented at the 24th Signal Processing and Communication Application Conference (SIU), Zonguldak, TURKEY, 2016, Accessed: 00, 2020. [Online]. Available: https://hdl.handle.net/11511/55707.