@ARTICLE{10.3389/fneur.2022.954142, AUTHOR={Shan, Zack Y. and Mohamed, Abdalla Z. and Andersen, Thu and Rendall, Shae and Kwiatek, Richard A. and Fante, Peter Del and Calhoun, Vince D. and Bhuta, Sandeep and Lagopoulos, Jim}, TITLE={Multimodal MRI of myalgic encephalomyelitis/chronic fatigue syndrome: A cross-sectional neuroimaging study toward its neuropathophysiology and diagnosis}, JOURNAL={Frontiers in Neurology}, VOLUME={13}, YEAR={2022}, URL={https://www.frontiersin.org/articles/10.3389/fneur.2022.954142}, DOI={10.3389/fneur.2022.954142}, ISSN={1664-2295}, ABSTRACT={IntroductionMyalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS), is a debilitating illness affecting up to 24 million people worldwide but concerningly there is no known mechanism for ME/CFS and no objective test for diagnosis. A series of our neuroimaging findings in ME/CFS, including functional MRI (fMRI) signal characteristics and structural changes in brain regions particularly sensitive to hypoxia, has informed the hypothesis that abnormal neurovascular coupling (NVC) may be the neurobiological origin of ME/CFS. NVC is a critical process for normal brain function, in which glutamate from an active neuron stimulates Ca2+ influx in adjacent neurons and astrocytes. In turn, increased Ca2+ concentrations in both astrocytes and neurons trigger the synthesis of vascular dilator factors to increase local blood flow assuring activated neurons are supplied with their energy needs.This study investigates NVC using multimodal MRIs: (1) hemodynamic response function (HRF) that represents regional brain blood flow changes in response to neural activities and will be modeled from a cognitive task fMRI; (2) respiration response function (RRF) represents autoregulation of regional blood flow due to carbon dioxide and will be modeled from breath-holding fMRI; (3) neural activity associated glutamate changes will be modeled from a cognitive task functional magnetic resonance spectroscopy. We also aim to develop a neuromarker for ME/CFS diagnosis by integrating the multimodal MRIs with a deep machine learning framework.Methods and analysisThis cross-sectional study will recruit 288 participants (91 ME/CFS, 61 individuals with chronic fatigue, 91 healthy controls with sedentary lifestyles, 45 fibromyalgia). The ME/CFS will be diagnosed by consensus diagnosis made by two clinicians using the Canadian Consensus Criteria 2003. Symptoms, vital signs, and activity measures will be collected alongside multimodal MRI.The HRF, RRF, and glutamate changes will be compared among four groups using one-way analysis of covariance (ANCOVA). Equivalent non-parametric methods will be used for measures that do not exhibit a normal distribution. The activity measure, body mass index, sex, age, depression, and anxiety will be included as covariates for all statistical analyses with the false discovery rate used to correct for multiple comparisons.The data will be randomly divided into a training (N = 188) and a validation (N = 100) group. Each MRI measure will be entered as input for a least absolute shrinkage and selection operator—regularized principal components regression to generate a brain pattern of distributed clusters that predict disease severity. The identified brain pattern will be integrated using multimodal deep Boltzmann machines as a neuromarker for predicting ME/CFS fatigue conditions. The receiver operating characteristic curve of the identified neuromarker will be determined using data from the validation group.Ethics and study registryThis study was reviewed and approved by University of the Sunshine Coast University Ethics committee (A191288) and has been registered with The Australian New Zealand Clinical Trials Registry (ACTRN12622001095752).Dissemination of resultsThe results will be disseminated through peer reviewed scientific manuscripts and conferences and to patients through social media and active engagement with ME/CFS associations.} }