Study of the Transfer Functions of Hippocampal Subfields during a Spatial Memory Task using High-Resolution fMRI
Xiaowei Zhuang1, Zhengshi Yang1, Tim Curran2, and Dietmar Cordes1,2

1Cleveland Clinic Lou Ruvo Center for Brain Health, Las Vegas, NV, United States, 2Department of Psychology and Neuroscience, University of Colorado Boulder, Boulder, CO, United States

Synopsis

In this abstract, the input/output transfer relationship in human hippocampal subfields (mainly CA1, CA3, and DG) was studied using fMRI during a spatial memory task with increments in the change of FOV in the stimuli. Whole brain activation was obtained for all lure v/s control contrasts. Mean activation t value for each hippocampal subregions (CA1, CA2&3 and CA4&DG) was extracted, averaged over all the subjects and plotted against FOV changes to compare with the existing models. K-means clustering was then applied. Data from one of the k-means clusters showed a pattern separation/completion curve similar to the animal model.

Purpose

The critical role of the hippocampus and nearby Medial Temporal Lobe (MTL) cortex in learning and memory is well documented1, 2, but the precise nature of the function of hippocampal subfields and the adjacent MTL cortical subregions remains unclear. Animal models have indicated an input/output transfer function for neural activity in hippocampal subfields3, which indicates neurons in the Cornu Ammonis area 1 (CA1) region respond linearly to the change of input, whereas neurons in Dentate Gyrus (DG) are more sensitive and respond non-linearly to small increment in the variation of input 4, 5. An fMRI study has been conducted to further investigate this input/output transfer relationship in human hippocampal subfields (mainly CA1, CA3 and DG) during a spatial memory task.

Methods

Imaging: fMRI was performed on 24 healthy subjects on a 3.0 T Siemens Trio scanner (32-channel head coil, multiband accelerate factor = 8, TR = 765ms, TE = 30 ms, flip angle = 44deg, FOV = 19.1 x 14.2 cm, 80 slices in oblique coronal orientation perpendicular to long axis of hippocampus, resolution = 1.65mm x 1.65mm x 2mm, BW = 1724 Hz/pixel, echo spacing = 0.72 ms and 2380 time frames). A repetition suppression spatial memory task was performed to examine differences in pattern separation and completion tendencies of hippocampal subfields, involving viewing of pairs of visual objects presented for a first time, repetitions of the same objects, or lures that were visually similar to the first-time objects. Three type of lures were presented, which differed from the original objects by a separation distance of 10% Field of View (FOV), 20% FOV, 30% FOV. A control task (pixelated background of the math task) occurred occasionally requiring subjects to perform simple addition. Analysis: All data were slice timing corrected, realigned to the co-planar SE-EPI image, distortion corrected using two SE-EPI acquisitions with reverse gradient polarity and smoothed with a 4mm Gaussian kernel. Time series were detrended after first 15 frames were eliminated and mass-univariate analysis was then applied to all 24 subjects. Whole brain t-statistics activation map for contrast Lure v/s Control were acquired and coregistered to the subject anatomical space. A standard MPRAGE T1 structural image was input into Freesurfer segmentation6 to get hippocampal subfields. Average t-statistics for new v/s control, repeat v/s control and lure v/s control with different FOV changes in CA1, CA3 and DG were plotted and extracted as a feature vector for each subject. K-means clustering was performed on the entire feature set for all 24 subjects. Average t-statistics for each cluster were plotted against the change of FOVs for different subregions to analyze the input/output transfer function in that subfield.

Results

Bilateral activations (p< 0.001, uncorrected (see Fig 1, whole brain activation map for lure with 20% change of FOV)) in hippocampus were observed for all contrast (lures v/s control). Mean t-value for all subregions, averaging over 24 subjects, indicates a higher activation in CA2&3 and CA4&DG regions than CA1 area (Fig 2A). Further, data from one of the k-means clusters (Fig 2B) showed patterns similar to the prediction model7.

Discussion

A pattern separation/completion curve similar to the animal model3, 7 is obtained in CA1 and DG regions with data from one of the k-means clusters. Activation in CA1 region shows a monotonic increase with the incremental change in FOV whereas in DG region, smaller increments in change in FOV result in faster activations and finally converge with CA1 region.

Conclusion

Our spatial memory fMRI design shows bilateral activations in major hippocampal regions with higher activations being observed in CA234&DG areas as compared to CA1 region. In addition, as the change of FOV increases, activations in all areas show an overall rising pattern and one cluster output from k-means indicates activity similar to pattern separation in the animal model. Future applications of our experimental design and method can be applied to monitor treatment effects on human memory disorders.

Acknowledgements

This research was supported by the NIH (grant number 7R01EB014284).

References

1. Fortin NJ et al., 2002, Nature Neuroscience. 2. Wallenstein GV et al., 1998, Trends in Neurosciences. 3. Stark CEL et al., 2011, Trends in Neurosciences. 4. Bakker A et al., 2008, Science. 5. Guzowski JF et al., 2004, Neuron. 6. Van Leemput K et al., 2009, Hippocampus. 7. Lacy JW et al., 2011, Learn Mem.

Figures

Figure1. Whole brain activation during memory related task; (t-statistic map for contrast: Lure (20% FOV) v/s Control; thresholded at t >3.5, p<0.001).

Figure2. A. Mean activation (t-statistics) for CA1, CA2&3 and CA4&DG, average over all subjects; B. Mean activation for subregions, average over one cluster output from k-means clustering. Activations in CA1 and DG regions show consistent patterns with model prediction.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
0485