The study was approved by the human ethics (HREC/15/QGC/63 and GU:2014/838) committee of Griffith University and the Gold Coast University Hospital where scanning was performed. Written informed consent was obtained from all individuals. 18 ME/CFS patients who met ICC criteria (Carruthers et al., 2011) and 26 age-matched healthy control subjects were recruited (see Table 1 for demographic information) through an online Lime survey. Furthermore, healthy controls and ME/CFS patients were excluded if they had an exclusionary medical disorder were: hyper/hypotensive, had an autoimmune dysfunction, attention deficit hyperactivity disorder, autoimmune disease, microvascular disease, or body mass index (BMI) > 35 or were pregnant or breastfeeding.

Clinical measures incorporated in cortical volume and thickness map regressions were collected as mentioned in Thapaliya et al. (2021). The 36-item SF36 short-form health survey questionnaire (Alonso et al., 1995), was completed by all subjects, and “Fatigue,” “SF36 physical (Phys_all)” and “SF36 mental scores (Ment_all)” were extracted. An “information processing score (Procinfo)” and a “Sleep disturbance score (SDS)” were obtained via a survey: “In the past month, how severe were the following symptoms (on a scale of 1–10, 1 being not a problem, 10 being extremely severe)” for symptoms “Difficulty processing information?” and “Sleep disturbances?” The “Heart rate (HR),” “Heart rate variability (HRV),” and “Respiratory rate (Resp)” were extracted from the power spectra of the pulse oximeter and respiration strap data recorded during a 15-min resting-state fMRI acquired in the same scanning session (“HR” and “Resp” from the frequency of the primary peak, and HRV from the full width at half maximum of the primary HR peak).

T1 weighted images for both ME/CFS and HC were acquired using a 3T Skyra MRI scanner (Siemens Healthcare, Erlangen, Germany) with a 64-channel head-neck coil (Nova Medical, Wilmington, NC, United States). Three-dimensional T1 weighted images were acquired using a T1 weighted magnetization prepared rapid gradient-echo (MPRAGE) sequence with a repetition time (TR) = 2,400 ms, echo time (TE) = 1.81 ms, flip-angle = 8°, acquisition matrix = 224 × 224 × 208, and voxel size 1 mm × 1 mm × 1 mm. The total acquisition time for T1w scans was 8:20 min:s.

FreeSurfer version 7.1.1 (Fischl, 2012) was run to generate cortical, sub-cortical volume and thickness from T1w images from ME/CFS patients and healthy controls using the Desikan Killiany parcelation scheme (Desikan et al., 2006). The default FreeSurfer command ‘‘recon-all’’ was run in a Macintosh computer (Operating system: Catalina, RAM = 36 GB, and core: 8). The ‘‘recon-all’’ processing includes motion correction, non-uniform intensity normalization, automated Talairach transformation, intensity normalization, removal of non-brain tissue, cortical parcelation, sub-cortical segmentation, gray and white matter boundary tessellation, automated topology correction, and surface deformation. Detailed information on the pipeline can be found here¹. Skull stripping and gray and white matter boundaries were checked visually, and participants were excluded if segmentation showed any error. The recon-all was performed using the “qcache” option and the analysis were performed using volume and thickness data with 10 mm full-width half maximum separately on the left and right hemisphere.

We performed group comparison of left and right hemisphere using a general linear model (GLM) (Fischl, 2012) by computing vertex-by-vertex for analysis of cortical volume and thickness using FreeSurfer. Individual structural maps were combined into a single dataset and resampled into MNI space using the FreeSurfer command “mris_preproc” (Fischl, 2012). GLM analysis was performed on the concatenated data of the left and right hemispheres using the FreeSurfer command “mri_glmfit” (Fischl, 2012). The multiple comparisons correction (cluster correction) (Hagler et al., 2006) was performed using “mri_glmfit-sim” (Fischl, 2012) with setting vertex-wise threshold at 1.3 and cluster-wise p-threshold of 0.05 to control for false positives.

We also performed cortical volume and thickness interaction-with-group regressions with clinical parameters to test for different relationships in ME/CFS and HC groups, that is, an abnormal relationship in ME/CFS. To perform the group interaction, we used the FreeSurfer GLM method by creating a FreeSurfer Group Descriptor (FSGD) file that describes a group of subjects and their accompanying data² and the contrast³. The design matrix is automatically created by FreeSurfer. The default method Different Offset Different Slopes (DODS) was used to perform group interaction in FreeSurfer. The “mri_glmfit” command was run with FSGD, and contrast and multiple comparison correction (cluster correction) was performed using FreeSurfer command line “mri_glmfit-sim.” The detail information about group interaction can be found in the given link https://surfer.nmr.mgh.harvard.edu/fswiki/FsTutorial/GroupAnalysis. The eight clinical parameters used as regressors were “HR,” “HRV,” “Phys_all,” “Procinfo,” “Ment_all,” “Resp,” and “SDS.” One ME/CFS patient was omitted from group interaction analysis due to missing clinical information (Procinfo, “Phys_all,” and “SDS”). ME/CFS patient data with clinical and autonomic measure outliers (one-“Procinfo,” one-“SDS,” and two-“Resp”) were also omitted from group interaction analysis.

Region-based statistical analysis was also performed on cortical and subcortical regions using SPSS version 27. All the statistical tests were controlled for age, gender, and total intracranial volume. Correction for multiple comparisons was implemented using false discovery rate (FDR).

We performed volumetric and thickness analysis on 18 ME/CFS patients and 26 HC. Figure 1 shows significant clusters with decreased volume in the left caudal middle frontal region (cluster size = 1,793 mm², p = 0.0016, X = −34.6; Y = 2.6, Z = 53.8) and decreased thickness in the right precuneus region (cluster size = 1,418 mm²; p = 0.013; X = 23.1, Y = −63.1, Z = 12.4).

We performed region-based analysis on the sub-cortical volume (Left and right: thalamus, caudate, putamen, pallidum, amygdala; posterior, anterior central regions of the corpus callosum; right, and total cortex volume) obtained directly from FreeSurfer as shown in Table 2. The central region of the corpus callosum, left and right hemisphere, and whole cortex volumes were significantly lower in ME/CFS compared to HC only before the multiple comparison correction (see Table 2). We only observed significantly larger volumes in left amygdala (p = 0.002) which survived the multiple comparison correction. The comparison of our significantly different volumetric regions in ME/CFS with previous findings are presented in Table 3.

Vertex-based interaction-with-group regressions were performed between cortical volume and thickness (left and right hemisphere) surface maps and eight clinical scores: “Fatigue,” “Phys_all,” “Ment_all,” “Procinfo,” “SDS,” “HR,” “HRV,” and “Resp.” Significant volume and/or thickness interaction-with-group regressions were detected for six regressors (“Fatigue,” “HRV,” “HR,” “SDS,” “Resp,” “Ment_all”). Volume and thickness clusters for which ME/CFS regression slopes significantly different to HC slopes are listed in Table 4.

Figure 2 shows four clusters with statistically significant volume or thickness interaction-with-group regressions with “Fatigue” and “HRV.” Fatigue showed a significantly different ME/CFS regressions in the right postcentral gyrus and inferior parietal lobe (see Figure 2, left). Cortical thickness interaction regressions with “HRV” showed significant clusters in the right superior parietal (Figure 2 left) and the left superior frontal gyrus (Figure 2 right). “HR” regressions showed significant clusters with abnormal volume and thickness in ME/CFS. The significant cluster of later occipitals (left and right) and caudal middle (right) frontal gyrus thickness and in the paracentral gyrus volume in the left hemisphere (see Figure 3). Cortical volume and thickness regression with “SDS” showed significant volume clusters in the left later occipital and superior frontal gyrus and significant thickness clusters in the right lingual gyrus (see Figure 4). Four significant volume and thickness clusters were detected in regression with “Resp” (see Figure 5). Significant volume cluster of left caudal middle frontal and right superior frontal gyrus and thickness cluster of left rostral middle frontal and superior frontal gyrus were abnormal in ME/CFS patients (see Figure 5). Cortical volume regression with “Ment_all” showed a significant cluster in the inferior parietal lobe of the right hemisphere (see Figure 5, right). The interaction-with group regression plot is shown in the Figure 6.

We detected significantly decreased volumes in the left caudal middle frontal cortex in ME/CFS patients. This region is involved in inhibition and modulation of attention (Japee et al., 2015) and participates in executive function (Andersson et al., 2009). A study of self-initiated elaborate encoding strategies (which rely on complex, highly effortful cognitive processes) demonstrated the involvement of left caudal middle frontal cortex (Husa et al., 2017). ME/CFS patients report memory and concentration problems, and difficulties in processing complex information (Jason et al., 1999) and perform worse than healthy controls in neuropsychological tests of attention, working memory, and processing speed (Marcel et al., 1996; Vercoulen et al., 1998). These deficits are consistent with the observed smaller left caudal middle frontal volume in ME/CFS.

Our ME/CFS patients also had reduced cortical thickness in the right precuneus which is involved in visual imagery, attention, and memory retrieval (Cavanna and Trimble, 2006). This is consistent with the ME/CFS symptom of difficulty in directing and maintaining visual attention (Hutchinson and Badham, 2013).

We also detected significant differences in the left amygdala volume in ME/CFS patients. The volume of the amygdala was significantly greater in ME/CFS which confirms an earlier VBM result (Finkelmeyer et al., 2018). Amygdala morphological changes can indicate a neuroinflammatory process (Lv et al., 2014; Nakatomi et al., 2014) or neuronal and synaptic alterations induced by stress (Roozendaal et al., 2009; Christoffel et al., 2011). Increased financial stress was associated with increased symptom severity in ME/CFS (Balinas et al., 2021) and better stress management skills lowered illness burden and fatigue severity in ME/CFS (Lattie et al., 2013). Increased amygdala volume in ME/CFS from exposure to stress may be mediated by the expression of Brain-derived neurotrophic factor (BDNF) (Bennett and Lagopoulos, 2014) which is altered in ME/CFS (Chen et al., 2008; Polli et al., 2020).

Vertex-based cortical volume and thickness interaction-with-group regressions with clinical measures yielded multiple significant clusters (Table 4 and Figures 2–5). In these clusters, regressions were oppositely directed for ME/CFS and HC, that is, ME/CFS regressions were abnormal (see Figure 6). We interpret inter-individual differences in local volume or thickness to be an expression of normal human variability. Figure 6 (x-values) shows this is similar for both ME/CFS and HC in the clusters illustrated. Insofar as volume or thickness is a surrogate for a functionally relevant feature such as myelination or axonal density, different correlations with clinical measures in a cluster indicate abnormal communication in ME/CFS within the control circuits that traverse the cluster and influence the clinical measure. This mechanism was proposed in an earlier MRI study of autonomic correlations (Barnden et al., 2016).

Cortical volume and thickness map interaction-with-group regressions with “Fatigue” and “Ment_all” scores both showed significant clusters in the inferior parietal lobe. The inferior parietal lobe is a hub of the default mode network (DMN) and the abnormal correlations detected here with fatigue and mental scores may be a manifestation of the same neuronal phenomenon that yielded diminished resting connectivity between inferior parietal and medial prefrontal DMN hubs in the same cohort (Shan et al., 2018).

We detected significant cortical thickness interaction regression with heart rate variability (HRV) in the left superior frontal gyrus. This is consistent with a resting-state functional MRI study which showed that HRV was positively correlated with BOLD activity in the superior frontal gyrus (Yoo et al., 2018).

We also demonstrated an abnormal correlation between respiratory rate (Resp) and cortical volume and thickness in the superior frontal gyrus, caudal middle frontal, and rostral middle frontal cortex. A pilot study in ME/CFS showed different respiratory rates in ME/CFS patients (Nijs et al., 2008). Our previous T1/T2 study also showed a group interaction with respiratory rate in the middle temporal gyrus, corpus callosum, and cerebral WM regions in ME/CFS patients (Thapaliya et al., 2020). A diffusion tensor imaging (DTI) study found an abnormal correlation between diffusion parameters correlation and “Resp” (Thapaliya et al., 2021) in the superior prefrontal cortex (BA 9) in ME/CFS patients.

Here we also detected abnormal cortical volume and thickness interaction-with-group regressions with “HR” in four cortical regions (Table 4), one in the middle frontal lobe. HR is faster in ME/CFS than controls in both supine and seated positions (Nelson et al., 2019). White matter (WM) volumes from voxel-based morphometry showed interaction-with-group regressions in bilateral prefrontal WM, hypothalamus and cerebellum (Barnden et al., 2016).

The autonomic measures HRV, Resp, and HR are regulated by the central autonomic network that involves the medial prefrontal cortex, insular cortex, amygdala, hypothalamus and midbrain, pons and medulla (Benarroch, 1993). Here the prefrontal cortex was involved in multiple interaction with group regressions with autonomic measures.

We also tested cortical volume and thickness maps for interaction-with group regressions with sleep disturbance score (SDS). Significant clusters were detected in the superior frontal, lingual and occipital cortex. Previous research on alcohol use disorder patients with sleep disorder showed reduced overall cortical volume (Wiers et al., 2015; Tomasi et al., 2019). Zhang et al. (2021) showed that longer sleep-wave and rapid eye movement (REM) sleep was significantly associated with greater cortical thickness. Diffusion tensor imaging showed abnormal inferior frontal gyrus correlations between “SDS” and DTI parameters in ME/CFS patients (Thapaliya et al., 2021). Another study using fMRI also showed activation of the inferior frontal gyrus after sleep deprivation (Vartanian et al., 2014). Thus, the clusters detected here do not agree with earlier “SDS” results and further study is required to resolve this difference.

The relatively small ME/CFS sample size will affect the power of the study to detect all the differences in cortical regions and their association with clinical measures. Larger populations should be investigated in future studies to ensure more accurate statistical results are obtained. The cortical volume and thickness are also affected by the choice of work station, operating system, processing software, and its version (Gronenschild et al., 2012; Perlaki et al., 2017; Seiger et al., 2018). Another limitation is that some of the clinical scores in this study were obtained by questionnaires, which by their subjective nature may limit interpretation of our findings. This study was a cross-sectional study. Longitudinal studies should be performed to test for progressive cortical volume and thickness changes in ME/CFS patients.