Prediction of cerebral blood flow using voxel-wise resting state functional MRI
Poster No:
442
Submission Type:
Abstract Submission
Authors:
Bhim Adhikari1, Hongjie Ke2, Yezhi Pan3, David Keator4, Daniel Amen5, Si Gao1, Paul Thompson6, Neda Jahanshad7, Jessica Turner8, Theo van Erp9, Mohammed Milad1, Jair Soares1, Vince Calhoun10, Juergen Dukart11, L. Elliot Hong1, Tianzhou Ma2, Peter Kochunov1
Institutions:
1UT Health Science Center at Houston, Houston, TX, 2University of Maryland at College Park, College Park, MD, 3University of Maryland School of Medicine, Baltimore, MD, 42Amen Clinics Inc., Costa Mesa, CA, 5Change Your Brain Change Your Life Foundation, Costa Mesa, CA, 6University of Southern California, Los Angeles, CA, 7University of Southern California,, Marina del Rey, CA, 8Wexner Medical Center, The Ohio State University, Columbus, OH, 9University of California, Irvine,, Irvine, CA, 10GSU/GATech/Emory, Atlanta, GA, 11Institute of Neuroscience and Medicine, Brain & Behaviour (INM‐7), Research Center Jülich, Jülich, Germany
First Author:
Co-Author(s):
Juergen Dukart
Institute of Neuroscience and Medicine, Brain & Behaviour (INM‐7), Research Center Jülich
Jülich, Germany
Institute of Neuroscience and Medicine, Brain & Behaviour (INM‐7), Research Center Jülich
Jülich, Germany
Introduction:
Regional cerebral blood flow (rCBF) is a putative biomarker for neuropsychiatric disorders including major depressive disorder (MDD). Here we show that rCBF can be predicted from resting-state functional MRI (rsfMRI) at voxel-level while correcting for partial volume averaging (PVA) artifacts. Cortical patterns of MDD-related CBF differences decoded from rsfMRI, using PVA-corrected approach showed excellent agreement with CBF measured using single photon emission computed tomography (SPECT) and arterial spin labeling (ASL).
Methods:
The Amish Connectome Project (ACP) consisted of N=300 participants with both rsfMRI and ASL data (age mean± s.d.: 37.5±16.3 years). We also analyzed the UK Biobank (UKBB) rsfMRI and 3D T1-weighted brain MRI data from N=2,290 participants with recurrent MDD (age mean±s.d.:62.1±7.4 years) and N=6,106 healthy controls (age mean±s.d.: 61.9±7.1 years). The Amen Clinics Inc. (ACI) data consisted of N=296 patients with recurrent or first episode MDD (age mean±s.d.:46.1±17.2) years) and N=76 healthy controls (age mean±s.d.: 42.2±17.2 years) with SPECT data. The resting-state analysis workflow developed by the Enhancing Neuro Imaging Genetics through Meta-Analysis (ENIGMA) consortium was used to process the rsfMRI data (Adhikari et al. 2018a, 2018b). CBF perfusion was estimated by using a standard single compartment ASL model. Partial volume effects correction was performed with a spatially regularized method (Chappell et al., 2011). T1-weighted image processing included registration to the MNI space, correction of B1 homogeneity, and removal of the non-brain tissue.
The support vector machine (SVM) method with a radial basis function kernel was used to predict voxel-wise CBF from rsfMRI data (Ding X, 2021). All statistical analyses were performed in R v4.3.2 (Yanase et al., 2005). Analyses were focused on evaluation of PVA correction for predicting voxel-wise CBF from rsfMRI data and were performed in two independent cohorts. In the ACP cohort, we evaluated the effects of PVA correction on the accuracy of CBF prediction. In UKBB, where only rsfMRI data was available, we used PVA-corrected CBF prediction to evaluate the pattern of regional CBF differences between MDD and controls.
The support vector machine (SVM) method with a radial basis function kernel was used to predict voxel-wise CBF from rsfMRI data (Ding X, 2021). All statistical analyses were performed in R v4.3.2 (Yanase et al., 2005). Analyses were focused on evaluation of PVA correction for predicting voxel-wise CBF from rsfMRI data and were performed in two independent cohorts. In the ACP cohort, we evaluated the effects of PVA correction on the accuracy of CBF prediction. In UKBB, where only rsfMRI data was available, we used PVA-corrected CBF prediction to evaluate the pattern of regional CBF differences between MDD and controls.
Results:
Voxel-wise CBF values were computed using the voxel-wise rsfMRI timeseries data with and without PVA correction. The PVA-corrected CBF predictions averaged across the whole brain showed numerically higher Pearson correlation with measured, whole-brain CBF values than uncorrected CBF predictions (r=0.68, p=2.5×10-13 vs. r=0.50, p=4.2×10-7, for corrected vs. original). The effect sizes (ESs) of the predicted CBF (for corrected model) differences between MDD and Healthy controls are shown in Figure 1 and Figure 2 (A). Stronger ESs were more negative and represented by hotter color. MDD was associated with significantly lower CBF for 32 cortical regions of interest. The strongest negative effect sizes were observed in superior frontal, inferior parietal and postcentral gyri (Cohen's d=-0.37±0.2, p<10-16). The regional effects for predicted CBF values in the UKBB showed statistically significant positive correlation with those computed from the ACI SPECT data (r=0.74, p=4.9×10-7 (Figure 2 (B)).
·Brain surface rendered effect size maps. Note, the stronger negative effect sizes are represented by hotter color.
·Cohen’s d effect size for predicted CBF with and without partial voxel averaging (PVA) correction and CBF measured by SPECT.
Conclusions:
This study demonstrated that rCBF can be predicted from the frequency spectrum of the rsfMRI data. This provides an opportunity to study effects of neuropsychiatric illnesses on rCBF because rsfMRI is much more widely available than ASL/SPECT/PET data in large and inclusive datasets such as the UKBB, ABCD and HCP, and ENIGMA Consortium clinical working groups. PVA-corrected rCBF value showed better agreement with ASL-derived CBF values in the testing dataset and SPECT-derived MDD ESs. The uncorrected rCBF values showed statistically weaker associations with ASL-derived CBF values and SPECT-derived MDD ESs. Overall, this study posits the rsfMRI-derived rCBF values as a putative practical and accurate biomarker for brain disorders.
Disorders of the Nervous System:
Psychiatric (eg. Depression, Anxiety, Schizophrenia) 1
Novel Imaging Acquisition Methods:
Multi-Modal Imaging 2
Keywords:
Cerebral Blood Flow
Single Photon Emission Computed Tomography (SPECT)
Other - partial volume correction; rsfMRI; support vector machine
1|2Indicates the priority used for review
Abstract Information
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Are you Internal Review Board (IRB) certified? Please note: Failure to have IRB, if applicable will lead to automatic rejection of abstract.
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
For human MRI, what field strength scanner do you use?
Which processing packages did you use for your study?
Provide references using APA citation style.
2. Adhikari B.M. et al. (2018b). Comparison of heritability estimates on resting state fMRI connectivity phenotypes using the ENIGMA analysis pipeline. Hum Brain Mapp, 39(12), 4893-4902.
3. Chappell, M. A. et al. (2011). Partial volume correction of multiple inversion time arterial spin labeling MRI data. Magn Reson Med 65, 1173-1183.
4. Ding X, L. J., Yang F, Cao J. (2021). Random radial basis function kernel-based support vector machine. Journal of the Franklin Institute 358(6), 10121-10140.
5. Meyer, D. et al. (2023). e1071: Misc Functions of the Department of Statistics, Probability Theory Group, (Formerly: E1071), TU Wien R package version 1.7-14.
6. Yanase, D. et al. (2005). Brain FDG PET study of normal aging in Japanese: effect of atrophy correction. Eur. J. Nucl. Med. Mol. Imaging, 32(7), 794–805.