Reduced functional connectivity in autoimmune encephalitis is explained by BOLD pattern incongruency
Poster No:
1732
Submission Type:
Abstract Submission
Authors:
Amy Romanello1,2, Stephan Krohn1,2, Carsten Finke1,2
Institutions:
1Charité - Universitätsmedizin Berlin, Berlin, Germany, 2Berlin School of Mind and Brain, Humboldt-Universität zu Berlin, Berlin, Germany
First Author:
Amy Romanello
Charité - Universitätsmedizin Berlin|Berlin School of Mind and Brain, Humboldt-Universität zu Berlin
Berlin, Germany|Berlin, Germany
Charité - Universitätsmedizin Berlin|Berlin School of Mind and Brain, Humboldt-Universität zu Berlin
Berlin, Germany|Berlin, Germany
Co-Author(s):
Stephan Krohn
Charité - Universitätsmedizin Berlin|Berlin School of Mind and Brain, Humboldt-Universität zu Berlin
Berlin, Germany|Berlin, Germany
Charité - Universitätsmedizin Berlin|Berlin School of Mind and Brain, Humboldt-Universität zu Berlin
Berlin, Germany|Berlin, Germany
Carsten Finke
Charité - Universitätsmedizin Berlin|Berlin School of Mind and Brain, Humboldt-Universität zu Berlin
Berlin, Germany|Berlin, Germany
Charité - Universitätsmedizin Berlin|Berlin School of Mind and Brain, Humboldt-Universität zu Berlin
Berlin, Germany|Berlin, Germany
Introduction:
Patients with anti-N-methyl-D-aspartate receptor encephalitis (NMDARE) present with a complex neuropsychiatric syndrome, yet clinical MRI sequences often show no abnormalities. Interestingly, resting-state functional MRI (rs-fMRI) studies in NMDARE have consistently found reduced functional connectivity (FC) between the hippocampus and default mode network (DMN). Given the mathematical definition of FC, an explanation for these findings must be present in the BOLD patterns themselves. However, it remains unclear what underlying signal property drives these FC alterations.
Against this background, we here leverage two recently introduced methodologies - a time-resolved signal complexity approach as in Krohn et al. (2023) combined with the edge time-series framework as in Esfahlani et al. (2020) - to compute a novel, pairwise measure that quantifies differences in BOLD-signal pattern distributions, yielding an "index of pattern incongruency" (IPI).
Against this background, we here leverage two recently introduced methodologies - a time-resolved signal complexity approach as in Krohn et al. (2023) combined with the edge time-series framework as in Esfahlani et al. (2020) - to compute a novel, pairwise measure that quantifies differences in BOLD-signal pattern distributions, yielding an "index of pattern incongruency" (IPI).
Methods:
Resting-state fMRI data from 75 healthy controls (HCs) and 75 patients with NMDARE were preprocessed and parcellated into 244 regions of interest (ROIs) using the Human Brainnetome Atlas. Static FC was calculated as the product-moment correlation coefficient between every pair of ROI time-series. Time-resolved signal complexity was calculated as weighted permutation entropy (WPE) with a sliding-window approach, resulting in a set of 244 ROI-wise complexity time-series for each participant. WPE estimates the irregularity of a time-series based on a symbolic encoding framework that is sensitive to both amplitude information and frequency content of a signal. Importantly, WPE is agnostic to which individual pattern drives the signal and instead computes Shannon entropy on the distribution across all possible patterns. Thus, it is possible for two different pattern distributions to have equal WPE values.
The calculation of IPI rests first on the application of the edge time-series framework to compute the moment-to-moment synchronization between each pair of complexity time-series, resulting in a continuous measure of complexity cofluctuation at the resolution of windows. We then isolated the windows with the highest 10% complexity cofluctuation, within which we examined the individual BOLD pattern distributions that underlie the WPE calculation. IPI was then calculated as the Euclidean distance between the pattern frequency distributions of two ROI time-series in each window. Metrics were calculated on the regional level with the hippocampus as well as the global level as distributions over all within- and between-network connections. ROI-wise between group differences were calculated using permutation-based T-tests and group differences in distributions using Wilcoxon rank-sum tests.
The calculation of IPI rests first on the application of the edge time-series framework to compute the moment-to-moment synchronization between each pair of complexity time-series, resulting in a continuous measure of complexity cofluctuation at the resolution of windows. We then isolated the windows with the highest 10% complexity cofluctuation, within which we examined the individual BOLD pattern distributions that underlie the WPE calculation. IPI was then calculated as the Euclidean distance between the pattern frequency distributions of two ROI time-series in each window. Metrics were calculated on the regional level with the hippocampus as well as the global level as distributions over all within- and between-network connections. ROI-wise between group differences were calculated using permutation-based T-tests and group differences in distributions using Wilcoxon rank-sum tests.
Results:
In line with previous work, we found that patients with NMDARE show significantly reduced FC between the hippocampus and several DMN regions, including medial frontal and cingulate cortices (all p <0.05). Analysis of BOLD-signal patterns revealed that decreases in FC are explained by higher IPI in patients (correlation of FC and IPI test statistics: rho = -0.451, p < 0.001). In between-network connections, patients had significantly reduced FC (Z = 14.40, p < 0.001), increased complexity cofluctuation (Z = -17.90, p < 0.001), and increased IPI (Z = -27.69, p <0.001). Within-network analyses revealed a similar trend with weaker effects: patients trended towards reduced FC (Z = 1.01, p = 0.31) and showed significantly increased complexity co-fluctuation (Z = -4.50, p < 0.001) and increased IPI (Z = -2.52, p = 0.01).
Conclusions:
Reduced functional connectivity between the hippocampus and DMN in NMDARE is explained by an increased incongruency of the underlying symbolic patterns in individual BOLD signals. The novel IPI metric links the covariance structure of BOLD signals to the underlying pattern variability within each signal and is sensitive to differences in functional brain dynamics that are undetected by WPE alone.
Disorders of the Nervous System:
Psychiatric (eg. Depression, Anxiety, Schizophrenia) 2
Modeling and Analysis Methods:
fMRI Connectivity and Network Modeling 1
Methods Development
Task-Independent and Resting-State Analysis
Novel Imaging Acquisition Methods:
BOLD fMRI
Keywords:
Computational Neuroscience
DISORDERS
FUNCTIONAL MRI
Other - Autoimmune encephalitis, BOLD-signal complexity
1|2Indicates the priority used for review
Provide references using author date format
Fadlallah, B. (2013), “Weighted-Permutation Entropy: A Complexity Measure for Time Series Incorporating Amplitude Information”, Physical Review E 87 (2): 022911.
Fan, L. (2016), “The Human Brainnetome Atlas: A New Brain Atlas Based on Connectional Architecture”, Cerebral Cortex 26 (8): 3508–26.
Finke, C. (2013), “Functional and Structural Brain Changes in Anti-N-Methyl-D-Aspartate Receptor Encephalitis”, Annals of Neurology, Vol 73 No 6, June
Krohn, S. (2023), “A Spatiotemporal Complexity Architecture of Human Brain Activity”, Science Advances 9 (5).