3876

Comparison of 3T and 7T PCC seed based spontaneous co-activation analysis
Wanyong Shin1, Crawford Anna1, Xiao Liu2,3, and Mark J Lowe1

1Radiology, Cleveland Clinic, Cleveland, OH, United States, 2Biomedical Engineering, Pennsylvania State University, University Park, PA, United States, 3Institute for Cyber Science, Pennsylvania State University, University Park, PA, United States

Synopsis

Dynamic functional connectivity techniques are increasingly being used investigate changes in neuronal network during resting state fMRI scanning. The spontaneous co-activation (CAP) approach has been proposed to reduce the dimensionality and shows the potential to characterize dynamic brain network in various ways. This study presents the first 7T CAP analysis, compared CAP analyses at 7T to 3T.

Introduction

Dynamic functional connectivity techniques are increasingly being used investigate changes in neuronal network during resting state fMRI scanning. The spontaneous co-activation (CAP) approach has been proposed to reduce the dimensionality (1) and shows the potential to characterize dynamic brain network in various ways (2). This study presents the first 7T CAP analysis, compared CAP analyses at 7T to 3T.

Method

Twenty eight healthy controls were scanned at 3T using single band EPI (TR=2.8s, 2x2x4mm3, 132 repetitions) and 18 healthy controls were scanned at 7T using simultaneous multi-slice (SMS) excited EPI sequence (TR=2.8s, MB factor=3, 1.2x1.2x1.5mm3, 128 repetitions). rsfMRI data was acquired simultaneously with pulse plethysmograph and respiratory belt recording. After the first 4 volumes were removed, physiologic noise was corrected using RETROICOR (3). Head motion was corrected using 3d volume registration and slice-wise motion correction (4). 3T data is normalized to MNI space using linear registration and resampled with isotropic 3mm voxel size. 7T data is aligned to a typical subject template in MNI space using non-linear transformation (5), and resampled with isotropic 1.5mm voxel size.

The PCC seed was selected as a sphere at with 6mm radius at MNI coordinate [0,-53,26]. To improve SNR, the time frames at the top 15% of seed signals were selected and the mask of the top 10% and the bottom 5% of voxels were chosen before k-mean clustering in both data set as described in (1). Spontaneous CAP patterns were calculated using kmeans (10,000 iteration) with 8 cluster number. The classified patterns were sorted from the largest to the smallest fraction [refs] and the consistency of each pattern was calculated by average CC between the cluster and each corresponding frame.

Result

PCC related rsfMRI network was decomposed to 8 CAPs from both 3T and 7T data sets at Z = −21, −9, 3, 15, 27, 39, and 51, respectively. Fig1 shows that cluster 1, 2 4, 5 and 7 resemble the default mode network, and cluster3 and 8 represent the visual sensory and motor network at 3T, respectively. Fig2 presents the default mode network in cluster 1, 2, 5, 6 and 7 and the visual sensory and motor networks in cluster 3 and 4 at 7T.

We found that 7T CAPs show the higher averaged z-scores in middle frontal gyrus, superior frontal gyrus, caudate nucleus, and hippocampus and the lower average z-score in insula, intraparietal sulcus in MFG, SFG and CN in Default mode network than 3T CAPs, as visualized with the warmer and the cooler color in Fig2.

Discussion

We observed that 7T CAP analysis captured the dynamic change of default network patterns with the high spatial resolution. The delineated caudate neucleus, hippocampus and thalamus were observed in 7T CAP analysis while thalamus was not clearly found in 3T CAP. Even with the smaller voxel size, e.g. 2.34 mm3 vs 16mm3, 7T CAP analysis was feasible in the relative small number of subjects.

Non-linear normalization step that was only used with 7T date could contribute to the outcome of 7T CAP analysis, but this study does not provide the source of the 7T CAP benefit.

Acknowledgements

Authors gratefully acknowledge technical support by Siemens Medical Solutions.

References

1. Liu X, Duyn JH. Time-varying functional network information extracted from brief instances of spontaneous brain activity. Proceedings of the National Academy of Sciences of the United States of America. 2013;110(11):4392-7.

2. Chen JE, Chang C, Greicius MD, Glover GH. Introducing co-activation pattern metrics to quantify spontaneous brain network dynamics. Neuroimage. 2015;111:476-88.

3. Glover GH, Li TQ, Ress D. Image-based method for retrospective correction of physiological motion effects in fMRI: RETROICOR. Magn Reson Med. 2000;44(1):162-7.

4. Beall EB, Lowe MJ. SimPACE: generating simulated motion corrupted BOLD data with synthetic-navigated acquisition for the development and evaluation of SLOMOCO: a new, highly effective slicewise motion correction. Neuroimage. 2014;101:21-34.

5. Avants BB, Tustison NJ, Song G, Cook PA, Klein A, Gee JC. A reproducible evaluation of ANTs similarity metric performance in brain image registration. Neuroimage. 2011;54(3):2033-44.

Figures

Fig1. Eight PCC seed co-activation patterns in 3T data (N=28). The colored areas indicate z-score within clustering (uncorrected p < 0.01), scaled from -1 (Blue) to 1 (Red). Fraction and consistency of each CAP are described in the right column.

Fig2. Eight PCC seed co-activation patterns in 7T data (N=18). The second left column presents the corresponding cluster in 3T with Pearson correlation coefficient. The colored areas indicate z-score within clustering (uncorrected p < 0.01), scaled from -1 (Blue) to 1 (Red). Fraction and consistency of each CAP are described in the right column.

Proc. Intl. Soc. Mag. Reson. Med. 27 (2019)
3876