Analysis of a distributed neural system involved in spatial information, novelty, and memory processing

Abstract

Perceiving a complex visual scene and encoding it into memory involves a hierarchical distributed network of brain regions, most notably the hippocampus (HIPP), parahippocampal gyrus (PHG), lingual gyrus (LNG), and inferior frontal gyrus (IFG). Lesion and imaging studies in humans have suggested that these regions are involved in spatial information processing as well as novelty and memory encoding; however, the relative contributions of these regions of interest (ROIs) are poorly understood. This study investigated regional dissociations in spatial information and novelty processing in the context of memory encoding using a 2 x 2 factorial design with factors Novelty (novel vs. repeated) and Stimulus (viewing scenes with rich vs. poor spatial information). Greater activation was observed in the right than left hemisphere; however, hemispheric effects did not differ across regions, novelty, or stimulus type. Significant novelty effects were observed in all four regions. A significant ROI x Stimulus interaction was observed - spatial information processing effects were largest effects in the LNG, significant in the PHG and HIPP and nonsignificant in the IFG. Novelty processing was stimulus dependent in the LNG and stimulus independent in the PHG, HIPP, and IFG. Analysis of the profile of Novelty x Stimulus interaction across ROIs provided evidence for a hierarchical independence in novelty processing characterized by increased dissociation from spatial information processing. Despite these differences in spatial information processing, memory performance for novel scenes with rich and poor spatial information was not significantly different. Memory performance was inversely correlated with right IFG activation, suggesting the involvement of this region in strategically flawed encoding effort. Stepwise regression analysis revealed that memory encoding accounted for only a small fraction of the variance (< 16%) in medial temporal lobe activation. The implications of these results for spatial information, novelty, and memory processing in each stage of the distributed network are discussed.

Figure 1

Examples of outdoor scenes with rich and poor spatial information (RSI and PSI) used in the study. RSI scenes (top row) contained several identifiable objects that bore a distinct spatial relation and scale to each other (landmarks). PSI scenes (bottom row) contained objects—e.g., mountains and clouds—that provided little information about the relative positions and scale of the objects with respect to each other.

Figure 2

Regions of interest (ROIs) were marked on average, Talairach‐normalized, high‐resolution structural images obtained from the 20 subjects used in the present study. (A) Left and right hemisphere lingual gyrus (shown in yellow), parahippocampal gyrus (shown in blue) and hippocampus (shown in red). (B) Inferior (shown in red) and middle frontal gyri (shown in blue). The protocol used to demarcate these structures is described in the text.

Figure 3

Composite map of activation across 20 subjects, in the lingual gyrus, parahippocampal gyrus, hippocampus, and inferior prefrontal gyrus, to visual scenes that were rich in spatial information (RSI). The activation height of each voxel represents the number of subjects who activated that voxel. This map is shown for illustrative purposes. Statistical analysis of activation across conditions and subjects was conducted using an ANOVA on the percentage of voxels activated by each subject in each ROI.

Figure 4

Plot of means showing significant three‐way ROI × NOVELTY × STIMULUS interaction. There was significant NOVELTY × STIMULUS interaction in the LNG but not in the HIPP, PHG, or IFG. Note the increasing trend towards nonsignificant interaction from the PHG to the HIPP. No interaction effects or main effects of spatial information processing were seen in the IFG.

Figure 5

Inverse correlation between right IFG activation and overall memory performance as indexed by Dprime for scenes with rich (RSI, top) and poor (PSI, bottom) spatial information.