A machine learning investigation of functional connectivity alterations in Post-COVID19 fatigue
Poster No:
576
Submission Type:
Late-Breaking Abstract Submission
Authors:
Daniel Martins1, Ottavia Dipasquale1, Brandi Eiff1, Silvia Rota1, Daniel van Hamelen1, Timothy Nicholson2, Mattia Veronese3, Adam Hampshire1, Federico Turkheimer21, Fernando Zelaya1, Steven Williams1
Institutions:
1Department of Neuroimaging, Institute of Psychiatry,Psychology & Neuroscience, King’s College London, London, United Kingdom, 2IoPPN, KCL, London, United Kingdom, 3University of Padua, Padova, Italy
First Author:
Daniel Martins, MD, PhD
Department of Neuroimaging, Institute of Psychiatry,Psychology & Neuroscience, King’s College London
London, United Kingdom
Department of Neuroimaging, Institute of Psychiatry,Psychology & Neuroscience, King’s College London
London, United Kingdom
Co-Author(s):
Ottavia Dipasquale
Department of Neuroimaging, Institute of Psychiatry,Psychology & Neuroscience, King’s College London
London, United Kingdom
Department of Neuroimaging, Institute of Psychiatry,Psychology & Neuroscience, King’s College London
London, United Kingdom
Brandi Eiff
Department of Neuroimaging, Institute of Psychiatry,Psychology & Neuroscience, King’s College London
London, United Kingdom
Department of Neuroimaging, Institute of Psychiatry,Psychology & Neuroscience, King’s College London
London, United Kingdom
Silvia Rota
Department of Neuroimaging, Institute of Psychiatry,Psychology & Neuroscience, King’s College London
London, United Kingdom
Department of Neuroimaging, Institute of Psychiatry,Psychology & Neuroscience, King’s College London
London, United Kingdom
Daniel van Hamelen
Department of Neuroimaging, Institute of Psychiatry,Psychology & Neuroscience, King’s College London
London, United Kingdom
Department of Neuroimaging, Institute of Psychiatry,Psychology & Neuroscience, King’s College London
London, United Kingdom
Adam Hampshire
Department of Neuroimaging, Institute of Psychiatry,Psychology & Neuroscience, King’s College London
London, United Kingdom
Department of Neuroimaging, Institute of Psychiatry,Psychology & Neuroscience, King’s College London
London, United Kingdom
Federico Turkheimer2
Department of Neuroimaging, Institute of Psychiatry,Psychology & Neuroscience, King’s College London
London, United Kingdom
Department of Neuroimaging, Institute of Psychiatry,Psychology & Neuroscience, King’s College London
London, United Kingdom
Fernando Zelaya
Department of Neuroimaging, Institute of Psychiatry,Psychology & Neuroscience, King’s College London
London, United Kingdom
Department of Neuroimaging, Institute of Psychiatry,Psychology & Neuroscience, King’s College London
London, United Kingdom
Steven Williams
Department of Neuroimaging, Institute of Psychiatry,Psychology & Neuroscience, King’s College London
London, United Kingdom
Department of Neuroimaging, Institute of Psychiatry,Psychology & Neuroscience, King’s College London
London, United Kingdom
Introduction:
Post-COVID-19 syndrome (PCS) is characterized by persistent fatigue, cognitive deficits and neuropsychiatric symptoms, yet the underlying neural mechanisms remain poorly understood. To address this gap, we leveraged multi-echo resting-state functional MRI (ME-rs-fMRI) combined with machine learning to investigate functional connectivity alterations differentiating PCS patients from fully recovered individuals.
Methods:
We conducted ME-rs-fMRI on 40 participants, comprising 20 PCS patients (mean age = 40.6 ± 10.9 years) and 20 fully recovered controls (mean age = 40.0 ± 10.4 years). The groups were matched for age, gender, and BMI. Data were preprocessed using a standard pipeline, which included denoising using the data-driven TEDANA pipeline to boost SNR. Functional connectivity was computed by extracting BOLD time-series from 83 cortical and subcortical regions defined by the Desikan-Killiany atlas, followed by Pearson correlation to generate connectivity matrices. To account for potential confounders, we regressed out the effects of age, gender, handedness (dexterity), and total intracranial volume (TIV). Dimensionality reduction was performed using principal component analysis (PCA), retaining 34 components which cumulatively explained 100% of the variance. A linear support vector machine (SVM) classifier was then trained to distinguish PCS patients from recovered controls based on these PCA components. Model performance was evaluated using five-fold stratified cross-validation, with 20% of the sample reserved for testing. To enhance neurobiological interpretability, we identified the top three PCA components with the highest SVM feature weights and analyzed their mean loadings across the seven Yeo intrinsic networks. Statistical significance of classification performance was assessed via 1,000 permutation tests, where group labels were shuffled to generate null distributions.
Results:
The SVM classification model achieved a cross-validated accuracy of 72.5% (pperm = 0.014), an F1 score of 0.6548 (pperm = 0.047), and a mean ROC-AUC of 0.76 (pperm = 0.027), demonstrating robust discrimination between PCS patients and recovered controls. The top three PCA components (10, 20, 28) contributing to classification explained 3.37%, 2.10%, and 1.44% of the variance, respectively. These components showed maximal PCA loadings in edges involving the limbic, somatomotor, and ventral attentional networks, which are known to support affective, sensorimotor, and cognitive functions.
Conclusions:
Functional connectivity alterations in the limbic, somatomotor, and attentional networks may underlie the persistent affective, cognitive, and sensorimotor symptoms reported in PCS. Despite the modest sample size, our machine learning approach reliably distinguished PCS patients from recovered individuals, indicating that functional connectivity biomarkers could potentially support patient stratification and personalized treatment planning. Future research with larger cohorts and longitudinal designs is needed to establish causal relationships, validate these findings, and explore the clinical utility of functional connectivity biomarkers for guiding targeted therapeutic interventions.
Disorders of the Nervous System:
Psychiatric (eg. Depression, Anxiety, Schizophrenia) 1
Modeling and Analysis Methods:
Classification and Predictive Modeling 2
Keywords:
FUNCTIONAL MRI
Machine Learning
Multivariate
Other - Post-COVID19 syndrome
1|2Indicates the priority used for review
·Figure 1: SVM Classification Performance: ROC curves illustrate 5-fold cross-validation performance. The table reports cross-validated accuracy, F1 score, and ROC-AUC, with significance confirmed via
·Figure 2: Top Features Driving SVM Classification. Brain maps show PCA loadings for the three components with the highest SVM feature weights, with red indicating positive and blue indicating negative
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?
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.
Taquet, M., Lonergan, M., & Harrison, P. J. (2021). Diabetes, neurological and psychiatric disorders risk after COVID-19. The Lancet Diabetes & Endocrinology, 9(7), 413–415. https://doi.org/10.1016/S2213-8587(21)00055-4
Luft, A. R., Pereda, E., & Banissy, M. J. (2022). Network neuroscience and machine learning in human neuroscience research: Promises and pitfalls in PCS research. Nature Neuroscience, 25(4), 523–533. https://doi.org/10.1038/s41593-022-01026-2
Hampshire, A., Chatfield, D. A., Manktelow, A. E., Jolly, A. E., Mazibuko, N., Grant, J. E., Patrick, F., Trender, W., Vivekananda, U., Chamberlain, S. R., & Fallon, S. J. (2022). Multivariate profile analysis of long COVID cognitive symptoms. Frontiers in Aging Neuroscience, 14, 850455. https://doi.org/10.3389/fnagi.2022.850455
Dallan, L. L., Stroud, M., Margolis, M., Zou, J., Chan, P., Pollak, T., & Frey, B. N. (2023). Resting-state imaging biomarkers of long COVID reveal persistent network dysfunctions. Brain, 146(1), 34–45. https://doi.org/10.1093/brain/awac455