Esin Karahan

Tensor Analysis and Fusion of Multimodal Brain Images

Current high-throughput data acquisition technologies probe dynamical systems with different imaging modalities, generating massive data sets at different spatial and temporal resolutions-posing challenging problems in multimodal data fusion. A case in point is the attempt to parse out the brain structures and networks that underpin human cognitive processes by analysis of different neuroimaging modalities (functional MRI, EEG, NIRS, etc.). We emphasize that the multimodal, multiscale nature of neuroimaging data is well reflected by a multiway (tensor) structure where the underlying processes can be summarized by a relatively small number of components or “atoms.” We introduce Markov-Penrose diagrams-an integration of Bayesian DAG and tensor network notation in order to analyze these models. These diagrams not only clarify matrix and tensor EEG and fMRI time/frequency analysis and inverse problems, but also help understand multimodal fusion via multiway partial least squares and coupled matrix-tensor factorization. We show here, for the first time, that Granger causal analysis of brain networks is a tensor regression problem, thus allowing the atomic decomposition of brain networks. Analysis of EEG and fMRI recordings shows the potential of the methods and suggests their use in other scientific domains.

Simultaneous EEG/fMRI Analysis of the Resonance Phenomena in Steady-State Visual Evoked Responses:

The stability of the steady-state visual evoked potentials (SSVEPs) across trials and subjects makes them a suitable tool for the investigation of the visual system. The reproducible pattern of the frequency characteristics of SSVEPs shows a global amplitude maximum around 10 Hz and additional local maxima around 20 and 40 Hz, which have been argued to represent resonant behavior of damped neuronal oscillators. Simultaneous electroencephalogram/functional magnetic resonance imaging (EEG/fMRI) measurement allows testing of the resonance hypothesis about the frequency-selective increases in SSVEP amplitudes in human subjects, because the total synaptic activity that is represented in the fMRI-Blood Oxygen Level Dependent (fMRI-BOLD) response would not increase but get synchronized at the resonance frequency. For this purpose, 40 healthy volunteers were visually stimulated with flickering light at systematically varying frequencies between 6 and 46 Hz, and the correlations between SSVEP amplitudes and the BOLD responses were computed. The SSVEP frequency characteristics of all subjects showed 3 frequency ranges with an amplitude maximum in each of them, which roughly correspond to alpha, beta and gamma bands of the EEG. The correlation maps between BOLD responses and SSVEP amplitude changes across the different stimulation frequencies within each frequency band showed no significant correlation in the alpha range, while significant correlations were obtained in the primary visual area for the beta and gamma bands. This non-linear relationship between the surface recorded SSVEP amplitudes and the BOLD responses of the visual cortex at stimulation frequencies around the alpha band supports the view that a resonance at the tuning frequency of the thalamo-cortical alpha oscillator in the visual system is responsible for the global amplitude maximum of the SSVEP around 10 Hz. Information gained from the SSVEP/fMRI analyses in the present study might be extrapolated to the EEG/fMRI analysis of the transient event-related potentials (ERPs) in terms of expecting more reliable and consistent correlations between EEG and fMRI responses, when the analyses are carried out on evoked or induced oscillations (spectral perturbations) in separate frequency bands instead of the time-domain ERP peaks.

Achromatic temporal-frequency responses of human lateral geniculate nucleus and primary visual cortex

The sensitivity of the sensory systems to temporal changes of the environment constitutes one of the critical issues in perception. In the present study, we investigated the human early visual systems dependency on the temporal frequency of visual input using fMRI. Blood oxygen level-dependent (BOLD) responses of the lateral geniculate nucleus (LGN) and primary visual cortex (V1) were investigated in a wide frequency range (6-46Hz) with fine frequency sampling (13 frequencies). Subject-specific functional-anatomic ROIs were derived from the combination of the anatomic template masks and the functional maps derived from multi-session fMRI analyses across all 13 stimulation conditions. Using functional-anatomic ROIs, average responses of LGN and V1 were calculated for each frequency. The V1 surface area was further parsed into 7 eccentricity sectors to detail central and peripheral responses. LGNs response revealed fluctuations on a background of non-significant decrease of the BOLD response with increasing stimulation frequency, while V1 response displayed similar fluctuations with a global maximum in the range of 8-12Hz, but a rapid and significant decrease with increasing stimulation frequency especially above 14Hz. This behavior of V1 response valid for both central and peripheral vision emphasizes that the profound low-pass effect of the visual system to visual input emerges in V1, presumably generated by the intra-cortical circuitry of V1 or projections from extra-striate areas. Besides, the high correlation between LGN and V1 BOLD responses across all visual stimulation frequencies supports the oscillatory tuning in thalamo-cortical interactions as previously claimed in electrophysiological studies.

Structural and Functional Evaluation of Glaucomatous Neurodegeneration from Eye to Visual Cortex using 1.5T MR Imaging: A Pilot Study

Aim: Glaucoma is an optic neuropathy and glaucomateus damage proceeds from retinal ganglion cells to brain. A better understanding of retrobulbar damage will enable us to develop more efficient strategies and a more accurate understanding of glaucoma. We evaluated retrobulbar glaucomatous damage with favorable techniques for 1.5T MR imaging.Material and methods: Five glaucoma cases and one healthy subject are included. Diffusion tensor MR imaging and functional MR images were taken with 1.5T MR. Correlation of optic nerve and corpus geniculatum laterale diffusion tensor MR parameters with eye findings were statistically evaluated.Results: Optic nerve damage and cortical hypofunction were shown with diffusion tensor MR and functional MR imaging, respectively. Correlations of the apparent diffusion coefficient with mean deviation, pattern standard deviation, retinal nerve fiber layer thickness in distal optic nerve and fractional anisotropy with ganglion cell counting in proximal optic nerves and correlations of retinal nerve fiber layer thickness with axial diffusivities in both ipsilateral and contralateral corpus geniculatum laterales and with fractional anisotropy in ipsilateral corpus geniculatum laterale were statistically significant.Conclusion: The eye-brain connection in glaucoma can be evaluated with routine clinical instruments. Our results also revealed the eye-to-visual-pathway integrity of glaucomatous neurodegeneration, which must be verified in larger series.

Mapping of the visual cortex: A Freesurfer™-based approach

Retinotopy is a tool for mapping of the visual cortex which provides detailed information about the correspondence between the visual field and its cortical representation in the visual subject. Therefore, the studies about retinotopic mapping is increasing rapidly as a branch of neuro science. Our study includes three main parts: The first one is the acquiring the data. While subject is stimulated with flickering and slowly moving checkerboard, data are acquired with 3T MR machine. In the second part, analysis part, data are analyzed and visualized with Freesurfer™. Since the procedures on Freesurfer™ are complicated and depends on high technical information, in the third part, we create a MATLAB™-based GUI to make these processes simpler.

Steady state visual evoked potential informed fMRI analysis for alpha, beta and gamma bands

In this study, simultaneous EEG and fMRI measurements are recorded from 18 healthy subjects during steady state visual stimulation. The stimulation frequencies that belong to the alpha, beta and gamma bands, at which the highest EEG amplitudes were observed, were determined for each subjects. The fMRI responses at those specific stimulation frequencies were anlayzed. Both EEG amplitudes and the fMRI parameters, namely the extent of the active area and the mean percentage positive BOLD change r were observed to be the highest at the alpha band and the lowest at the gamma band.

Multivariate classification of fMRI images

Functional Magnetic Resonance Imaging (fMRI) gives vast amount of information on the neural activity of the brain. Researchers analyse fMRI data to investigate the functions and structure of the brain. Using machine learning tools that have been widely used in recent years in fMRI area, has enabled to predict the cognitive states of subject which is called “brain reading” also. In this study, fMRI data is taken from Ishai et. al. who explored the representation of object categories in brain. By using Naive Gauss Classifier and Support Vector Machines, it is tried to identify the patterns of objects and by analyzing fMRG images prediction on the cognitive state of the subject is performed.