Imaging the human medial temporal lobe with high-resolution fMRI

Abstract

High-resolution functional MRI (hr-fMRI) affords unique leverage on the functional properties of human medial temporal lobe (MTL) substructures. We review initial hr-fMRI efforts to delineate (1) encoding and retrieval processes within the hippocampal circuit, (2) hippocampal subfield contributions to pattern separation and pattern completion, and (3) the representational capabilities of distinct MTL subregions. Extant data reveal functional heterogeneity within human MTL and highlight the promise of hr-fMRI for bridging human, animal, and computational approaches to understanding MTL function.

Figure 1. Subfield analysis methods

(a) High-resolution anatomical images of human MTL displaying manually-drawn hippocampal ROIs: CA₁ (green), DG/CA_2/3 (blue), and the subiculum (red), as well as PRC (pink), ERC (cyan) and PHC (yellow). (b) Group subfield activity projected onto a flattened representation of the MTL demonstrating greater CA₁ involvement in navigating a virtual town from multiple (MSP) relative to single starting points (SSP) (modified from Suthana et al., 2009). (c) Group-level statistical maps overlaid on a representative participant's anatomical images demonstrating performance-related probe period activity on a delayed-match-to-sample task in posterior hippocampal and parahippocampal regions (modified from Olsen et al., 2009).

Figure 2. Hippocampal encoding and retrieval effects

(a) Subfield activity projected onto a flattened representation of the MTL demonstrating encoding activity in DG/CA_2/3 and retrieval activity in the subiculum. Flatmap is accompanied by bar graphs indicating averaged activity in DG/CA_2/3 and the subiculum across four consecutive blocks in which participants encoded and recalled the same face-name associations (modified from Zeineh et al., 2003). (b) Averaged subicular activity in response to faces and scenes presented once or repeatedly. Stimuli presented once were subsequently tested outside the scanner, with memory performance used to back-sort encoding trials as recognized with high confidence, low confidence, or missed (modified from Preston et al., 2010).

Figure 3. Pattern separation and delay-period activity

(a) Sample stimuli illustrating original and lure versions of objects. Activation clusters (white overlay) in DG/CA_2/3 (red) and CA₁ (blue) demonstrating biases toward pattern separation and completion, respectively. Bar graphs represent averaged activity extracted from each functional ROI in response to first presentation, repeat, and lure trials (modified from Bakker et al., 2008). (b) Timecourse activity extracted from anatomically defined regions of interest demonstrating delay period activity correlated with memory performance. Bar graphs depict mean signal change during study, delay, and probe (modified from Olsen et al., 2009).

Figure 4. Multivariate pattern analysis in the MTL

Scattergrams depict the accuracy with which objects were differentiated from scenes (left) and different classes of objects from one-another (right) in the hippocampus, PHC and PRC using MVPA classifiers trained and tested on the data of individual participants (colored squares). Chance performance is indicated by the dashed line (modified from Diana et al., 2008).