Neural Substrates of Visual Spatial and Non-spatial Top-down Attentional Process; an MVPA study

Abstract

Introduction:

Top-down attention is the cognitive process to concentrate selectively on the specific feature of information with internal guidance. Top-down attention can be categorized into two distinct types; Spatial attention and Non-spatial attention, handling non-spatial information such as colors and shapes. Many studies have investigated the common substrates for the top-down attention in general, however, each types of attention may be processed differently considering the is still unclear. In this study, our aim was to identify the neural substrates involved in coding each types of attention. Methods:

Subjects Nine healthy subjects (4 females), ages 20~28 years old, were recruited. During the pre-session, information about the experiment was provided to induce the effective attentional process in the main experiment.

Experiment procedure Visual spatial (left and right) and non-spatial (black and white) stimuli were presented. Cue of attention was presented for 1s after 6s cross fixation period. Cues were 'Left' and 'Right' for spatial attention task, and 'High' and 'Low' for non-spatial attention task. Top-down attention period for 2~4s was given before the task. Task is the time-gap detection to judge whether the target was flickering with short (50ms) or long (200ms) time gap at cued location or cued color. Then the subject was asked to press the response pad accordingly with right index or middle fingers. There were 5 sessions and each session had 48 trials.

MRI acquisition and pre-processing Imaging was acquired by a 3T Siemens Verio scanner with 32-channels. T1-weighted (TR = 2800ms, TE = 14ms) and functional T2*-weighted images (TR = 2000ms, TE = 35ms) were taken. Pre-processing was processed by Statistical Parametric Mapping (SPM) 12 in MATLAB.

fMRI GLM analysis According to the type of attention, General linear model (GLM) analysis was performed. Top-down attention period was analyzed by block model design. There were 4 regressors for 4 conditions of attention (left, right, high, and low). fMRI MVPA analysis We used support vector machine for multivariate pattern analysis (MVPA). MVPA searchlight was performed for all voxels in grey matter with 27 voxel searchlight volume (cube). Five SVM kernels (linear, polynomial, quadratic, multil-ayer perceptron, Gaussian radial basis function) were used for calculating max classification accuracy. Results:

First, GLM analysis was performed to find the common and different neural substrates according to the types of attention. There were common neural substrates for visual spatial and non-spatial attention, but no different substrates between types of attention were found. The substrates were left lingual gyrus, bilateral inferior parietal lobe, left postcentral gyrus, supplementary motor area and bilateral insula. Second, MVPA searchlight was performed to find the regions which show different pattern of activation according to the types of attention. Bilateral superior frontal regions (SFG) and right paracentral region were identified that can discriminate the visual spatial and non-spatial attention. In addition, the patterns of left medial superior frontal and right superior occipital, calcarine area can discriminate the visual spatial left and right attention, while those of cuneus and left middle occipital region can discriminate the visual color attention (white, black). Conclusions:

By using MVPA searchlight approach, we could discriminate the brain regions that process visual spatial and non-spatial information, that were not identified through GLM analysis. The activation pattern in SFG can discriminate spatial and non-spatial attention. The activation patterns of some other association cortices can discriminate the specific feature further. We conclude that SFG may be the top region which processes the initial discrimination of visual spatial and non-spatial features during attentional process.