Right Arcuate Fasciculus Abnormality in Chronic Fatigue Syndrome1

Note: This copy is for your personal non-commercial use only. To order presentation-ready copies for distribution to your colleagues or clients, contact us at www.rsna.org/rsnarights. ORIGINAL R

Right Arcuate Fasciculus ESEARCH

Abnormality in Chronic Fatigue n

Syndrome1 NEURORADIOLOGY

Michael M. Zeineh, MD, PhD Purpose: To identify whether patients with chronic fatigue syndrome James Kang, MD (CFS) have differences in gross brain structure, micro- Scott W. Atlas, MD scopic structure, or brain perfusion that may explain their Mira M. Raman, MS symptoms. Allan L. Reiss, MD Jane L. Norris, PA Materials and Fifteen patients with CFS were identified by means of Ian Valencia, BS Methods: retrospective review with an institutional review board– Jose G. Montoya, MD approved waiver of consent and waiver of authorization. Fourteen age- and sex-matched control subjects provided informed consent in accordance with the institutional review board and HIPAA. All subjects underwent 3.0-T volumetric T1-weighted magnetic resonance (MR) imaging, with two diffusion-tensor imaging (DTI) acquisitions and arterial spin labeling (ASL). Open source software was used to segment supratentorial gray and white matter and cerebrospinal fluid to compare gray and white matter volumes and cortical thickness. DTI data were processed with automated fiber quantification, which was used to compare piecewise fractional anisotropy (FA) along 20 tracks. For the volumetric analysis, a regression was performed to account for differences in age, handedness, and total intracranial volume, and for the DTI, FA was compared piecewise along tracks by using an unpaired t test. The open source software segmentation was used to compare cerebral blood flow as measured with ASL.

Results: In the CFS population, FA was increased in the right arcuate fasciculus (P = .0015), and in right-handers, FA was also increased in the right inferior longitudinal fasciculus (ILF) (P = .0008). In patients with CFS, right anterior arcuate FA increased with disease severity (r = 0.649, P = .026). Bilateral white matter volumes were reduced in CFS (mean 6 standard deviation, 467 581 mm3 6 47 610 for patients vs 504 864 mm3 6 68 126 for control subjects, P = .0026), and cortical thickness increased in both right arcuate end points, the middle temporal (T = 4.25) and precentral (T = 6.47) gyri, and one right ILF end point, the occipital lobe (T = 5.36). ASL showed no signiﬁcant differences.

1 From the Department of Radiology, Lucas Center for Conclusion: Bilateral white matter atrophy is present in CFS. No dif- Imaging, Stanford University School of Medicine, 1201 ferences in perfusion were noted. Right hemispheric in- Welch Rd, Room P271, Stanford, CA 94305-5488. Received creased FA may reflect degeneration of crossing fibers or May 7, 2014; revision requested June 10; revision received strengthening of short-range fibers. Right anterior arcuate July 4; accepted July 16; final version accepted July 29. FA may serve as a biomarker for CFS. Supported by the Division of Infectious Disease CFS Fund. M.M.Z. supported by GE Healthcare. Address correspon- q dence to M.M.Z. (e-mail: [email protected]). RSNA, 2014 q RSNA, 2014 Online supplemental material is available for this article.

Radiology: Volume 274: Number 2—February 2015 n radiology.rsna.org 517 NEURORADIOLOGY: Right Arcuate Fasciculus Abnormality in Chronic Fatigue Syndrome Zeineh et al

hronic fatigue syndrome (CFS) is has been explored across a variety of none of these patients were excluded. a debilitating disorder character- neurodegenerative disorders. White More than 300 control subjects from a Cized by 6 or more months of per- matter microstructure may be altered database of volunteers were examined sistent or relapsing fatigue without any in CFS as a consequence of inflamma- to find age-matched (within 1 year) associated medical or psychiatric disor- tion (12) or as part of the neurocog- and sex-matched participants without der (1). The high prevalence of 2–4 per nitive physiology of fatigue (13), but it any history of major depression, CFS 1000 people in the United States (2,3), has not been investigated to date with or chronic fatigue, or substance abuse combined with the profound disability DTI. in the past year; 28 eligible volunteers (4) and poor prognosis (5), motivates The purpose of this study was to (a) were contacted, and 14 chose to par- urgent scientific investigation. Brain identify differences in gross brain struc- ticipate and underwent MR imaging. imaging could aid in diagnosis and ture in CFS by using T1-weighted gray Control subjects provided informed prognosis. However, structural imaging and white matter volumetric analysis, consent in accordance with the insti- findings have been inconsistent: Three (b) detect microstructural abnormal- tutional review board and the Health voxel-based morphometry studies dem- ities underlying CFS by using DTI, and Insurance Portability and Accountabil- onstrated gray matter atrophy in dif- (c) detect global alterations in brain ity Act. The 20-item Multidimensional fering locations when specified (6–8), perfusion by using pseudocontinuous Fatigue Inventory (MFI-20) was ad- while one study in which cerebrospinal arterial spin labeling (ASL). ministered to each subject, which has fluid was segmented yielded no atrophy been validated as a reproducible in- (9). Brain perfusion studies have shown strument for the identification of CFS inconsistent decreases (10), with a Materials and Methods (14–17). The MFI-20 score is used to well-controlled study of monozygotic assess general, physical, and mental twins showing no differences (11). No Subjects fatigue, as well as reduced motivation technique has provided either a patho- This study began in June 2011, and and activity; higher MFI-20 scores in- physiological understanding of the dis- follow-up was completed in November dicate increased severity. Subjects order or served as a biomarker. 2013. Institutional review board ap- completed an expanded Edinburgh With diffusion-tensor imaging (DTI), proval with a waiver of consent and a handedness inventory (http://www. the random motion of water is used to waiver of authorization were obtained, brainmapping.org/shared/Edinburgh. demonstrate brain microstructure and and a retrospective review was con- php), with scores thresholded at 48 ducted from 2008 to the present for Advances in Knowledge patients evaluated in the university CFS n Patients with severe chronic clinic. Inclusion criteria were (a) the Published online before print fatigue syndrome (CFS) have clinical criteria for CFS (fatigue for a 10.1148/radiol.14141079 Content codes: duration of 6 months or longer, along increased fractional anisotropy Radiology 2015; 274:517–526 (FA) in the anterior right arcuate with having at least four of the eight Abbreviations: fasciculus when compared with Fukuda symptoms: impaired memory ASL = arterial spin labeling control subjects (P = .0015), with or concentration, sore throat, tender lymph nodes, headaches, muscle pain, CFS = chronic fatigue syndrome most right-handed patients CI = confidence interval joint pain, unrefreshing sleep, and having maximal FA in the ante- DTI = diffusion-tensor imaging rior 10% of the arcuate of higher postexertional malaise) and (b) ongoing FA = fractional anisotropy than 0.6, while in the same memory and concentration symptoms ILF = inferior longitudinal fasciculus MFI-20 = 20-item Multidimensional Fatigue Inventory region, most right-handed con- that cause a severe enough impairment that the physician determined clinical ROC = receiver operating characteristic trol subjects have an FA of less ROI = region of interest than 0.6. magnetic resonance (MR) imaging was appropriate to exclude other diagnoses. Author contributions: n The right hemisphere in younger All patients with CFS except one were Guarantors of integrity of entire study, M.M.Z., J.K., J.G.M.; patients with CFS exhibits evaluated by one clinician (J.G.M., with study concepts/study design or data acquisition or data regions of cortical thickening that 10 years of experience with CFS). No analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all are connected via the right ar- other inclusion or exclusion criteria authors; approval of final version of submitted manuscript, cuate fasciculus, specifically in were used for patients with CFS. Among the right middle temporal and all authors; agrees to ensure any questions related to the 259 charts reviewed, 15 patients were work are appropriately resolved, all authors; literature precentral gyri. identified who met these criteria, and research, M.M.Z., J.K., J.G.M.; clinical studies, M.M.Z., J.K., n White matter volumes are lower S.W.A., J.L.N., I.V., J.G.M.; experimental studies, M.M.Z., in patients with CFS compared J.K., A.L.R., I.V.; statistical analysis, M.M.Z., J.K., A.L.R.; with control subjects (467 581 Implication for Patient Care and manuscript editing, M.M.Z., J.K., S.W.A., A.L.R., I.V., J.G.M. mm3 6 47 610 vs 504 864 mm3 6 n Right anterior arcuate FA may be 68 126, P = .0026). a biomarker for CFS. Conflicts of interest are listed at the end of this article.

518 radiology.rsna.org n Radiology: Volume 274: Number 2—February 2015 NEURORADIOLOGY: Right Arcuate Fasciculus Abnormality in Chronic Fatigue Syndrome Zeineh et al

and higher. Four subjects declined this proton-density image, and imaging time volumes were excluded from analysis. inventory and simply described them- of 4 minutes 50 seconds). One patient Data were corrected for motion with a selves as left-handed, right-handed, or with CFS did not undergo ASL imaging rigid-body transformation and aligned ambidextrous. Left-handed and ambi- because of a technical error. to the b of 0 mm/sec2 image, which was dextrous subjects were pooled as non– then aligned to the T1-weighted vol- right-handed subjects. Segmentation: Volumetry ume in anterior commissure–posterior A FreeSurfer pipeline analysis con- commissure space. Automated fiber MR Imaging Acquisition ducted by using the T1-weighted im- quantification performs automated Subjects were imaged with one of two ages (http://surfer.nmr.mgh.harvard. tractography of 20 tracks, nine in each identical GE 3T 750HDx 60-cm–bore edu/) produced cortical, white mat- hemisphere (anterior thalamic radia- magnets (with scheduling dictating ter, and subcortical segmentations tions, corticospinal, cingulate cingulum, the imaging unit chosen) by using an (18). These segmentations underwent parahippocampal cingulum, inferior eight-channel phased-array receive- blinded manual editing (by J.K., with frontal occipital fasciculus, inferior lon- only head coil and body-transmit coil. 2 years of neuroradiology experience, gitudinal fasciculus [ILF], superior lon- For volumetric assessment, we per- and M.M.Z., with 17 years of segmen- gitudinal fasciculus, uncinate fasciculus, formed axial three-dimensional T1- tation and quantitative neuroimaging and arcuate fasciculus) and two bilat- weighted inversion-recovery spoiled experience) to improve skull stripping eral tracks (forceps major and minor gradient-echo and three-dimensional (typically, portions of the skull base, of the corpus callosum). Each tract (a) axial brain volume imaging with ar- dura, and dural venous sinuses are was cleaned of extraneous fibers that ray spatial sensitivity encoding tech- erroneously characterized as cortex), deviated too far from the fiber core and nique acceleration of 2, or ASSET 2, extend the white matter segmentation truncated at standard region-of-interest (repetition time msec/echo time msec, to include all subcortical white matter (ROI) positions and (b) was normal- 9.15/3.7; inversion time msec, 450; (due to residual B0 inhomogeneity af- ized to 100 pixels in length. With this flip angle, 13°; bandwidth, 25; field ter correction), and retract the white process, some tracks could be identi- of view, 240 mm; 256 3 256 matrix; matter segmentation when it seeps into fied in all subjects. 1.2-mm-thick sections; 130 sections the cortex (which occurs near primary acquired; imaging time, 2 minutes 50 somatosensory cortex at the vertex ASL Acquisition seconds). Two diffusion-weighted se- with its inherently reduced contrast ASL signal intensity was divided by sig- quences were performed, the first with with the subjacent white matter at T1- nal intensity in coplanar proton-density higher spatial resolution and the second weighted imaging) (19). An additional images and multiplied by a scaling fac- with higher angular resolution: (a) one step of blinded manual editing was per- tor to deliver maps of cerebral blood b of 0 mm/sec2 image acquired, 40 di- formed for quantification of subcortical flow in milligrams per milliliter per rections at b of 2000 mm/sec2, twice gray structures by using FreeView (by minute (21) (www.nmr.mgh.harvard. refocused, two-dimensional axial, AS- removing extra voxels from the hippo- edu/~jjchen/ASL.html). The proton- SET 2, frequency right/left, 8000/95.5, campal and thalamic segmentations). density image acquired as part of the field of view of 240 mm, 128 3 128 This produces a complete segmenta- ASL sequence was used to align with matrix reconstructed at 256 3 256, tion of the cortex, supratentorial white the T1-weighted volume by using the 2-mm-thick sections with 0-mm gap, 59 matter, deep gray nuclei, and ventri- FMRIB (Functional MRI of the Brain) sections acquired, two signals acquired, cles. For computing total intracranial Software Library, or FSL, “flirt” func- and imaging time of 11 minutes; (b) and volume, we summed the supratentorial tion with six degrees of freedom and one b of 0 mm/sec2 image acquired, 50 gray matter, supratentorial white mat- a normalized mutual information cost directions at b of 2000 mm/sec2, twice- ter, and cerebrospinal fluid compart- function, so cerebral blood flow images refocused two-dimensional axial, AS- ments (the latter was estimated from were in the same space as the Free- SET 2, frequency right/left, 8000/90.2, the sulcal, lateral, and third ventricular Surfer segmentations. This alignment field of view of 240 mm, 96 3 96 matrix and choroid plexus volumes). was manually inspected by M.M.Z. to reconstructed at 256 3 256, 4-mm- ensure accuracy. Cerebral blood flow thick sections with 0-mm gap, 36 sec- DTI Acquisition was averaged over the segmentations tions acquired, two signals acquired, The DTI acquisitions were imported of the cerebral cortex, supratentorial and imaging time of 13 minutes 44 separately, along with the T1-weighted white matter, basal ganglia, thalamus, seconds. Pseudocontinuous ASL was volume, into two separate and inde- and hippocampi. performed (three-dimensional axial pendent instances of automated fiber imaging, 4800/10.6, bandwidth of 62, quantification (http://white.stanford. Statistical Analysis field of view of 220, 128 3 128 matrix, edu/newlm/index.php/AFQ, performed All statistical analyses were performed 4-mm-thick sections, 63 sections ac- by M.M.Z.) (20). Imaging volumes were by M.M.Z. For all volumetric compar- quired, three signals acquired with inspected manually, and occasional ar- isons, a regression was computed by one signal acquired for the unlabeled tifactual diffusion-weighted imaging using age, total intracranial volume,

Radiology: Volume 274: Number 2—February 2015 n radiology.rsna.org 519 NEURORADIOLOGY: Right Arcuate Fasciculus Abnormality in Chronic Fatigue Syndrome Zeineh et al

disease, and handedness as indepen- based on regional differences in subject “fslstats” (25,26), and a spheroid was dent variables, a recommended proce- anatomy, as well as differences attrib- plotted with a volume equal to the clus- dure for these types of analyses (22), by utable to crossing fibers. A track had a ter volume by using MATLAB software using Stata software (StataCorp, Col- significant difference if the P value for (Mathworks, Natick, Mass). lege Station, Tex). Uncorrected P values the t test was below this threshold for For the ASL, unpaired t tests were for the comparison of the volumes of at least one point along the track. The used to compare cerebral blood flow the six regions evaluated—cortical gray high-spatial-resolution first DTI acquisi- for each of the regions between groups, matter, supratentorial white matter, tion was used to generate a hypothesis: with P values reported uncorrected. thalami, hippocampi, basal ganglia, and All tracks were tested for significant The results of all automated steps rostral middle frontal gyri—are report- differences of FA, correcting for multi- were verified by M.M.Z. either visually ed, along with the Bonferroni-correct- ple comparisons within each track, but or numerically for plausibility. All re- ed threshold for significance (.05/6 = not across all tracks. The high-angular- sults were interpreted with agreement .0083). The right and left white matter resolution second DTI acquisition was by M.M.Z., A.L.R, M.M.R, J.K., and compartments and middle frontal gyri used to test a hypothesis: Significant J.G.M. were evaluated separately by using the results from the hypothesis-generating same regression procedure. Exclusively data set were tested individually for within the CFS population, a Pearson significant differences in FA, correcting Results correlation coefficient was computed for multiple comparisons within each between regional volumes and MFI-20 track, and not needing a correction Clinical Data scores. across tracks because of the focused Thirteen of the 15 patients with CFS For the cortical thickness evalua- hypothesis testing. Maximal FA and and 10 of 14 control subjects were tion, we first made global comparisons mean axial diffusivity and radial diffu- right-handed. Women constituted 55% by examining the mean cortical thick- sivity were computed over the anterior of participants (eight patients with CFS ness across both hemispheres together, 10% of the right arcuate fasciculus. and eight control subjects), and 45% again comparing via regression. We This maximal FA was correlated with of the participants were men (seven then used the FreeSurfer “qdec” mod- the subject’s MFI-20 scores separately patients with CFS and six control sub- ule to compare cortical thickness be- for patients and control subjects. Ax- jects). Mean age 6 standard deviation tween the CFS group and the control ial diffusivity and radial diffusivity were was 46.2 years 6 14.2 (range, 20–66 group, regressing age as a covariate compared between groups by using years) and was statistically equivalent (22), and handedness as an additional an unpaired t test. A receiver oper- for women and men (46.6 years 6 discrete factor. By using a different off- ating characteristic (ROC) curve was 15.6 [range, 20–66 years] for women set and different slope for each group plotted separately for the hypothesis- vs 45.6 years 6 12.9 [range, 23–60 and smoothing of 10 mm, we corrected generating and hypothesis-testing data years] for men; P = .86). Mean ages of for multiple comparisons with a false- sets by identifying the accuracy of using the CFS and control groups were sta- discovery rate of .05. Cortical thickness maximal FA in the anterior 10% of the tistically equivalent (46.5 years 6 13.2 was regressed in the CFS group with right arcuate fasciculus to diagnose CFS and 46.6 years 6 14.6, respectively; P MFI-20 scores as an additional covari- across multiple FA thresholds (with = .98). MFI-20 scores were significantly ate, also by using “qdec.” the FA threshold spaced .05 apart) different from patients compared with For the DTI, fractional anisotropy by using Excel (Microsoft, Redmond, control subjects (MFI-20: 79.60 years (FA) was sampled along each identified Wash). The 95% confidence intervals 6 15.3 vs 27.64 years 6 8.54, respec- track and compared piecewise between (CIs) for sensitivity and specificity were tively; P , .001). The mean duration groups by using an unpaired t test to computed by using the Stata “diagti” of symptoms for CFS participants was compare both cohorts. Additional test- module, and the area under the ROC 12.1 years 6 6.9. ing was performed in right-handed curve was determined with the function individuals from each group because “roctab.” For combined visualization of Segmentation differences in language lateralization tractography and cortical thickness re- Volumetry.—For the entire cohort, can affect several tracks (including the sults, we first depicted the right arcuate there was significantly lower total su- right arcuate) (23). Automated fiber and right ILF in one subject by using pratentorial white matter volume for quantification uses a permutation- automated fiber quantification. Then, patients with CFS compared with con- based method to correct for multiple we drew ROIs around the three largest trol subjects, when accounting for age, comparisons between groups along a clusters of significantly different thick- total intracranial volume, and handed- single track (24), producing a thresh- nesses by using “qdec” and transformed ness (Table 1). This reduction was pre- old P value equivalent to a corrected P the ROIs into this single subject’s native sent bilaterally (left, P = .0035; right, value of .05. These P value thresholds space. To plot these ROIs alongside the P = .0033). Additionally, total thalamic vary per track, likely because tracks tractography, the center of mass for volume tended to be lower in CFS, but have unique correlation structures each cluster was extracted by using FSL this did not reach significance after

520 radiology.rsna.org n Radiology: Volume 274: Number 2—February 2015 NEURORADIOLOGY: Right Arcuate Fasciculus Abnormality in Chronic Fatigue Syndrome Zeineh et al

Table 1 Volumetry Comparison between Patients with CFS and Control Subjects and Correlation with MFI-20 Score P Value for Volume in Patients Volume in Control b Value for the Disease Status in r Value for MFI-20 Bonferroni-corrected Brain Region with CFS (mm3) Subjects (mm3) Fitted Slope Regression Analysis Scores in CFS P Values

Cortical gray matter 448 899 6 59 082 449 627 6 47 163 13 120 .0728 0.3658 .1800 Supratentorial white matter 467 581 6 47 610* 504 864 6 68 126* 217 439* .0026* 0.0214 .9396 Thalami 13 115 6 1870 14 430 6 1812 21009 .0469 0.2196 .4317 Hippocampi 7655 6 528 7950 6 781 2284 .1341 0.1256 .6555 Basal ganglia 18 982 6 2951 19 321 6 2832 72 .8934 0.5564 .0312 Rostral middle frontal gyri 30 323 6 5471 30 144 6 4733 1711 .1357 0.2343 .4006

Note.—r values demonstrate the correlations between CFS patient volumes and their corresponding MFI-20 scores. r values were calculated with the Pearson correlation coefﬁcient and are shown with the associated P values, with a Bonferroni-corrected threshold of .0083 (to control for the six regions tested). * Signiﬁcant difference.

correction for multiple comparisons. Table 2 Total cortical gray matter volume was statistically equivalent. Prefrontal cor- Regions of Significant Increases in Right-Hemispheric Cortical Thickness in Patients tex volumes (identified in other studies with CFS Compared with Control Subjects, Corrected for Multiple Comparisons of CFS, where foci are best matched by Location Area (mm2) T Score X Value Y Value Z Value the rostral middle frontal gyri as segmented with FreeSurfer) (6,27) were Lateral occipital 100.14 5.3592 25.3 288.4 29 also statistically equivalent between Precentral 89.19 5.3675 40.4 210.7 34 Middle temporal 35.84 4.2466 63.3 225.5 212.3 groups. However, the right middle fron- Postcentral 16.84 4.5786 61.4 211.9 17.2 tal gyrus trended toward increased vol- Pars orbitalis 8.83 4.0795 40.2 48.4 24.9 ume in the CFS group (15 446 mm3 6 2771 in the CFS group and 14 721 mm3 Note.—X, Y, and Z refer to Montreal Neurological Institute 305 coordinates. 6 2158 in control subjects; P = .0212). Within the CFS population, the MFI-20 score trended toward a positive correlation with basal ganglia volume (r = only significant difference was that fasciculus, pixel positions 90–100) and 0.5564, P = .0312). the anterior right arcuate fasciculus a decrease in radial diffusivity (0.364 Thickness.—Global cortical thick- (identified in 13 of 15 patients and 14 6 0.069 for patients vs 0.425 6 0.039 ness was equivalent between popula- of 14 control subjects) demonstrated for control subjects). The correlation tions (2.437 mm 6 0.102 in patients significantly higher FA in patients com- between maximal FA over the same with CFS vs 2.418 mm 6 0.091 in pared with control subjects, correcting portions of this track and MFI-20 in control subjects; P = .932) (Table 2, for multiple-comparisons within but the CFS population was not significant Fig 1). No focal left hemispheric differ- not across tracks (Table 3; Figs 2, 3). (r = 0.276, P = .4). Of note, the two ences were present. However, the right By performing the same unpaired t end points of the right arcuate fascic- hemisphere demonstrated five regions test just on the right arcuate fasciculus ulus are adjacent to the region of in- of increased cortical thickness when in the second hypothesis-testing DTI creased cortical thickening in the right accounting for age and handedness, data set (also identified in 13 of 15 pa- precentral and middle temporal gyri which was most notable in the youn- tients and 14 of 14 control subjects), (Figs 1, 3; Movie 1 [online]). ger, not the older, patients with CFS, we confirmed with a highly significant DTI in right-handed subjects with no regions of decreased cortical P value the same result by testing this only.—Right anterior arcuate FA was thickness (Table 2, Fig 1). No regions single hypothesis (P = .0001, signif- again significantly increased in CFS demonstrated cortical thickness that icant after correction within track). (Fig 2a, first DTI acquisition, P = increased significantly with age. No sig- This FA increase was accompanied by .0006; Fig 2b, second DTI acquisition, nificant correlations with MFI-20 were an increase in axial diffusivity ([1.117 P = .00036; both highly significant af- present bilaterally in the CFS group. 6 0.089] 3 1023 mm2/sec for patients ter correction within track). The cor- DTI in all subjects, right- and with CFS vs [1.005 6 0.084] 3 1023 relation between maximal FA over the left-handed.—In the high-resolution mm2/sec for control subjects, averaged anterior 10% of this track and MFI-20 hypothesis-generating data set, the over the anterior 10% of the arcuate in the CFS population was significant in

Radiology: Volume 274: Number 2—February 2015 n radiology.rsna.org 521 NEURORADIOLOGY: Right Arcuate Fasciculus Abnormality in Chronic Fatigue Syndrome Zeineh et al

Figure 1

Figure 1: (a) Plots of cortical thickness are shown according to age in the right occipital, precentral, and middle temporal gyral ROIs. (b) Superimposed on the inﬂated atlas brain image are regions of right-hemisphere increased cortical thickness in patients with CFS compared with control subjects after accounting for differences in age and handedness are highlighted in red (arrows). Blue = precentral, green = middle temporal, red = occipital, white = postcentral, and yellow = orbitofrontal. The right precentral region consists of two connected foci that FreeSurfer identiﬁed as one contiguous region.

hypothesis-testing data set, the same measurements were 72.7% (95% CI: 39.0%, 94.0%), 90.0% (95% CI: 55.5%, 99.7%), and 0.891 (95% CI: 0.696, 0.98) respectively.

ASL Findings ASL demonstrated no significant difference in perfusion to the cortex (672 mL per 100 mg per minute 6 123 for pa- the hypothesis-generating data set but right occipital region of cortical thick- tients with CFS vs 633 mL per 100 mg not the hypothesis-testing data set (r = ening (Figs 1, 3; Movie 1 [online]). per minute 6 145 for control subjects; 0.649 [P = .026] and r = 0.472 [P = An ROC curve was calculated by us- P = .39), supratentorial white matter .136], respectively). There was no cor- ing maximal FA in the anterior 10% of (495 mL per 100 mg per minute 6 88 relation between MFI-20 and the same the right arcuate fasciculus to diagnose for patients with CFS vs 496 mL per pixel locations in control subjects (r = CFS in right-handed patients (Fig 4 100 mg per minute 6 113 for control 20.02). FA was also increased in the shows the hypothesis-testing data subjects; P = .84), basal ganglia (523 right anterior ILF (Figs 2c, 2d, 3; first set ROC). A threshold of 0.6 would mL per 100 mg per minute 6 84 for pa- DTI acquisition, P = .0008; second DTI be used to correctly diagnose CFS tients with CFS vs 512 mL per 100 mg acquisition, P = .0018; both significant in right-handed patients in the hy- per minute 6 081 for control subjects; after correcting within track). However, pothesis-generating data set, with a P = .69), thalami (599 mL per 100 mg the peak FA was of greater magnitude sensitivity, specificity, and area un- per minute 6 116 for patients with CFS difference and more consistent in lo- der the ROC curve of 81.8% (95% vs 587 mL per 100 mg per minute 6 cation in the arcuate fasciculus com- CI: 48.2%, 97.7%), 100% (95% CI: 116 for control subjects; P = .78), or pared with the ILF. The occipital lobe 69.2%, 100%), and 0.918 (95% CI: hippocampi (587 mL per 100 mg per end point of the ILF was adjacent to the 0.696, 0.988), respectively. In the minute 6 106 for patients with CFS vs

522 radiology.rsna.org n Radiology: Volume 274: Number 2—February 2015 NEURORADIOLOGY: Right Arcuate Fasciculus Abnormality in Chronic Fatigue Syndrome Zeineh et al

Table 3 Statistical Tests for Signiﬁcance of Difference in FA across Tracks All Patients Right-Handers No. of Patients and No. of Patients and Control Subjects with Control Subjects with Track P Value P Value Threshold Track Present P Value P Value Threshold Track Present

Left thalamic radiation .2144 .0027 15, 14 .0891 .0022 13, 10 Right thalamic radiation .0360 .0019 15, 14 .0585 .0031 13, 10 Left corticospinal .0087 .0033 15, 14 .0050 .0026 13, 10 Right corticospinal .0499 .0024 15, 14 .0620 .0034 13, 10 Left cingulum cingulate .0029 .0012 15, 14 .0054 .0013 12, 10 Right cingulum cingulate .1325 .0019 15, 14 .2324 .0016 13, 10 Left cingulum hippocampus .0566 .0021 15, 14 .0979 .0023 13, 10 Right cingulum hippocampus .1107 .0027 14, 14 .0483 .0023 12, 10 Callosum forceps major .0854 .0016 13, 14 .0603 .0011 13, 10 Callosum forceps minor .1175 .0022 15, 14 .1404 .0022 13, 10 Left inferior fronto-occipital fasciculus .0133 .0014 15, 14 .0383 .0013 13, 10 Right inferior fronto-occipital fasciculus .0099 .0011 15, 14 .0285 .0015 11, 10 Left ILF .0041 .0026 15, 14 .0149 .0019 13, 10 Right ILF .0055 .0022 14, 14 .0008* .0015 13, 10 Left superior longitudinal fasciculus .0842 .0040 15, 14 .0861 .0045 13, 10 Right superior longitudinal fasciculus .1001 .0045 15, 14 .0873 .0046 13, 10 Left uncinate .2574 .0031 15, 14 .0622 .0025 13, 10 Right uncinate .2390 .0051 15, 14 .0606 .0034 13, 10 Left arcuate .0210 .0022 15, 14 .0343 .0016 13, 10 Right arcuate .0015* .0023 13, 14 .0006* .0028 11, 10

* P values are signiﬁcant after correcting for multiple comparisons within the track (ie, under the P value threshold).

mL per 100 mg per minute 6 118 for Additionally, two cortical regions con- of crossing fibers. Consistent with control subjects; P = .84). nected via the arcuate fasciculus ex- our study, others have reported that hibited increased thickness (28–30): the right arcuate fasciculus is some- the right middle temporal and precen- times not found with tractography, Discussion tral gyri. Similarly, in right-handers, in part because the arcuate is more This DTI study of CFS demonstrated the ILF showed that increased FA an- lateralized to the left hemisphere in increased FA in the right anterior teriorly increased thickness of a cor- right-handers (23,32). We found that arcuate fasciculus, and this increase responding right occipital region. Our arcuate differences between patients correlated with disease severity. An examination yielded no differences in with CFS and control subjects were automated, user-independent, micro- perfusion. most striking among right-handed structural DTI analysis identified this While microstructure has not been patients, suggesting that hemispheric increased FA and verified it in a second studied to date in CFS, increased FA differences with handedness and data set from the same subjects ana- is an unexpected finding for a disor- language (23) are an additional source lyzed independently. In right-handed der characterized by reduced cogni- of variance. patients with CFS, this increase was tive abilities. However, increased FA Our volumetry results stand in correlated with the patients’ MFI-20 has been previously reported in the contradistinction to the literature scores. Thus, in populations of CFS corona radiata in Alzheimer disease and show reduced gray matter vol- that have severe concentration and (31), possibly due to degeneration of umes, either globally or in the pre- memory problems, right anterior ar- crossing fibers. Given that the differ- frontal cortices (6,7,27,33). In fact, cuate FA may serve as a biomarker for ence in CFS only involved the ante- we found that the prefrontal volumes the disease. An analysis of volumetric rior 10% of the arcuate, we suspect trended toward being slightly higher data in the same subjects was used this local phenomenon could repre- in patients with CFS. Previous inves- to identify reduced bilateral supra- sent strengthening of fibers that join tigators have exclusively used voxel- tentorial white matter volumes, sug- a small portion of the anterior ar- based morphometry, which is based gesting a global white matter process. cuate, or alternatively a weakening on many factors other than cortical

Radiology: Volume 274: Number 2—February 2015 n radiology.rsna.org 523 NEURORADIOLOGY: Right Arcuate Fasciculus Abnormality in Chronic Fatigue Syndrome Zeineh et al

Figure 2

Figure 2: (a) FA is plotted along the right arcuate fasciculus from the first diffusion-weighted acquisition; anterior = 0, posterior = 100. Thin blue dotted lines correspond to 61 standard error. Only right-handers are included in this Figure. The blue arrow highlights the region of maximal increase of FA in patients with CFS compared with control subjects. (b) A plot is given for the second diffusion-weighted acquisition. (c, d) Similar plots are given for the right ILF, with the yellow arrow on c similarly highlighting the region of maximal increase of FA. thickness (34). The discordance be- well-controlled perfusion study of CFS between MFI-20 and right anterior tween the literature and our results involving monozygotic twins (11), sug- arcuate FA was only significant in the may be explained by our precise gesting perfusion is not affected in hypothesis-generating data set, not the methods, with a careful examination CFS. hypothesis-testing data set. Although of cortical architecture and correction Even though the hypothesis-testing this may be technically related to the for covariates, such as age and total data set confirms the veracity of in- reduced spatial resolution of the lat- intracranial volume (22). creased FA in the right anterior arcu- ter data set, the utility of arcuate FA Although we did not control for ate, it was determined in the same sub- as a disease biomarker cannot be caffeine exposure (35), our nega- jects and cannot be considered a fully conclusively confirmed on this study. tive ASL result is consistent with a independent data set. The correlation While the regions of increased cortical

524 radiology.rsna.org n Radiology: Volume 274: Number 2—February 2015 NEURORADIOLOGY: Right Arcuate Fasciculus Abnormality in Chronic Fatigue Syndrome Zeineh et al

Figure 3

Figure 3: Reconstructed MR image shows the right arcuate (blue tracks and arrows) and ILFs (yellow tracks and arrows) in a single representative subject. These two tracks are overlaid on their respective track proﬁles (the centroid of each track was averaged across subjects and depicted as the large tubular structures at the core of each track). The track proﬁle is colored according to the T score of track-based FA, showing that the maximal increase in FA is in the anterior arcuate and ILFs. The red, blue, and green spheres correspond to size and locations of increased cortical thickness from Figure 1 in the right occipital, precentral, and middle temporal regions, respectively. The green arrows also point to the middle temporal region of increased thickness.

Figure 4 networks with resting-state functional MR imaging.

Disclosures of Conflicts of Interest: M.M.Z. Activities related to the present article: disclosed no relevant relationships. Activities not related to the present article: GE Healthcare provided research funding to the institution that was unrelated to this study. Other relationships: disclosed no relevant relationships. J.K. disclosed no relevant relationships. S.W.A. disclosed no relevant relationships. M.M.R. disclosed no relevant relationships. A.L.R. disclosed no relevant relationships. J.L.N. disclosed no relevant relationships. I.V. disclosed no relevant relationships. J.G.M. disclosed no relevant relationships. Figure 4: ROC curve is shown for the diagnosis of CFS by measuring the maximal FA in the anterior 10% of the right arcuate fasciculus in right-handers in the hypothesis-testing data set (second DTI acquisition). The number next to References each data point indicates the FA threshold for that point in the ROC curve. 1. Fukuda K, Straus SE, Hickie I, Sharpe MC, Dobbins JG, Komaroff A. The chronic fatigue syndrome: a comprehensive approach to its definition and study. Inter- thickness are in close proximity to Future work includes validating national Chronic Fatigue Syndrome Study the end points of the arcuate and ILF, this finding in a larger right-handed co- Group. Ann Intern Med 1994;121(12): these foci are relatively small and im- hort, teasing apart the components of 953–959. perfectly aligned, suggesting a small crossing and short-range fibers in the 2. Reyes M, Nisenbaum R, Hoaglin DC, et al. effect size. Overall, this study has a anterior arcuate by using a multiple- Prevalence and incidence of chronic fatigue small number of subjects, so all of the shell (multiple b value) acquisition and syndrome in Wichita, Kansas. Arch Intern findings in this study require replica- examining the time-course in a longitu- Med 2003;163(13):1530–1536. tion and exploration in a larger group dinal study, possibly with interventions, 3. Jason LA, Richman JA, Rademaker AW, of subjects. and investigating right-hemisphere et al. A community-based study of chronic

Radiology: Volume 274: Number 2—February 2015 n radiology.rsna.org 525 NEURORADIOLOGY: Right Arcuate Fasciculus Abnormality in Chronic Fatigue Syndrome Zeineh et al

fatigue syndrome. Arch Intern Med (MFI) psychometric qualities of an instru- 26. Smith SM, Jenkinson M, Woolrich MW, et 1999;159(18):2129–2137. ment to assess fatigue. J Psychosom Res al. Advances in functional and structural MR 1995;39(3):315–325. image analysis and implementation as FSL. 4. Natelson BH, Johnson SK, DeLuca J, et Neuroimage 2004;23(Suppl 1):S208–S219. al. Reducing heterogeneity in chronic fa- 15. Gentile S, Delarozière JC, Favre F, Sam- tigue syndrome: a comparison with depres- buc R, San Marco JL. Validation of the 27. de Lange FP, Koers A, Kalkman JS, et sion and multiple sclerosis. Clin Infect Dis French ‘multidimensional fatigue inven- al. Increase in prefrontal cortical volume 1995;21(5):1204–1210. tory’ (MFI 20). Eur J Cancer Care (Engl) following cognitive behavioural therapy in 2003;12(1):58–64. patients with chronic fatigue syndrome. 5. Cairns R, Hotopf M. A systematic re- Brain 2008;131(Pt 8):2172–2180. view describing the prognosis of chronic 16. Reeves WC, Wagner D, Nisenbaum R, et al. fatigue syndrome. Occup Med (Lond) Chronic fatigue syndrome—a clinically em- 28. Phillips OR, Clark KA, Woods RP, et al. To- 2005;55(1):20–31. pirical approach to its definition and study. pographical relationships between arcuate BMC Med 2005;3:19. fasciculus connectivity and cortical thickness. 6. Okada T, Tanaka M, Kuratsune H, Wata- Hum Brain Mapp 2011;32(11):1788–1801. nabe Y, Sadato N. Mechanisms underlying 17. Lin JM, Brimmer DJ, Maloney EM, Nyarko fatigue: a voxel-based morphometric study E, Belue R, Reeves WC. Further validation 29. Catani M, Mesulam M. The arcuate fascicu- of chronic fatigue syndrome. BMC Neurol of the Multidimensional Fatigue Inventory in lus and the disconnection theme in language 2004;4(1):14. a US adult population sample. Popul Health and aphasia: history and current state. Cor- Metr 2009;7:18. tex 2008;44(8):953–961. 7. de Lange FP, Kalkman JS, Bleijenberg G, Ha- goort P, van der Meer JW, Toni I. Gray mat- 18. Reuter M, Rosas HD, Fischl B. Highly ac- 30. Yeatman JD, Dougherty RF, Rykhlevskaia ter volume reduction in the chronic fatigue curate inverse consistent registration: a E, et al. Anatomical properties of the ar- syndrome. Neuroimage 2005;26(3):777–781. robust approach. Neuroimage 2010;53(4): cuate fasciculus predict phonological and 1181–1196. reading skills in children. J Cogn Neurosci 8. Puri BK, Jakeman PM, Agour M, et al. 2011;23(11):3304–3317. Regional grey and white matter volumet- 19. Steen RG, Reddick WE, Ogg RJ. More than ric changes in myalgic encephalomyelitis meets the eye: significant regional hetero- 31. Douaud G, Jbabdi S, Behrens TE, et al. DTI (chronic fatigue syndrome): a voxel-based geneity in human cortical T1. Magn Reson measures in crossing-fibre areas: increased morphometry 3 T MRI study. Br J Radiol Imaging 2000;18(4):361–368. diffusion anisotropy reveals early white mat- 2012;85(1015):e270–e273. ter alteration in MCI and mild Alzheimer’s 20. Yeatman JD, Dougherty RF, Myall NJ, disease. Neuroimage 2011;55(3):880–890. 9. Perrin R, Embleton K, Pentreath VW, Jack- Wandell BA, Feldman HM. Tract pro- son A. Longitudinal MRI shows no cerebral files of white matter properties: automat- 32. Catani M, Allin MP, Husain M, et al. Sym- abnormality in chronic fatigue syndrome. Br ing fiber-tract quantification. PLoS ONE metries in human brain language pathways J Radiol 2010;83(989):419–423. 2012;7(11):e49790. correlate with verbal recall. Proc Natl Acad Sci U S A 2007;104(43):17163–17168. 10. Biswal B, Kunwar P, Natelson BH. Cerebral 21. Dai W, Garcia D, de Bazelaire C, Alsop blood flow is reduced in chronic fatigue DC. Continuous flow-driven inversion for 33. Puri BK, Counsell SJ, Zaman R, et al. Rela- syndrome as assessed by arterial spin label- arterial spin labeling using pulsed radio tive increase in choline in the occipital cor- ing. J Neurol Sci 2011;301(1-2):9–11. frequency and gradient fields. Magn Reson tex in chronic fatigue syndrome. Acta Psy- Med 2008;60(6):1488–1497. chiatr Scand 2002;106(3):224–226. 11. Lewis DH, Mayberg HS, Fischer ME, et al. Monozygotic twins discordant for chronic fa- 22. Barnes J, Ridgway GR, Bartlett J, et al. 34. Hutton C, Draganski B, Ashburner J, Weis- tigue syndrome: regional cerebral blood flow Head size, age and gender adjustment in kopf N. A comparison between voxel-based SPECT. Radiology 2001;219(3):766–773. MRI studies: a necessary nuisance? Neuro- cortical thickness and voxel-based mor- image 2010;53(4):1244–1255. phometry in normal aging. Neuroimage 12. Morris G, Maes M. Myalgic encephalomyeli- 2009;48(2):371–380. tis/chronic fatigue syndrome and encephalo- 23. Häberling IS, Badzakova-Trajkov G, Cor- myelitis disseminata/multiple sclerosis show ballis MC. Asymmetries of the arcuate 35. Wang DJ, Chen Y, Fernández-Seara MA, remarkable levels of similarity in phenom- fasciculus in monozygotic twins: genetic Detre JA. Potentials and challenges for ar- enology and neuroimmune characteristics. and nongenetic influences. PLoS ONE terial spin labeling in pharmacological mag- BMC Med 2013;11:205. 2013;8(1):e52315. netic resonance imaging. J Pharmacol Exp Ther 2011;337(2):359–366. 13. Genova HM, Rajagopalan V, Deluca J, et al. 24. Nichols TE, Holmes AP. Nonparametric per- Examination of cognitive fatigue in multiple mutation tests for functional neuroimaging: sclerosis using functional magnetic reso- a primer with examples. Hum Brain Mapp nance imaging and diffusion tensor imaging. 2002;15(1):1–25. PLoS ONE 2013;8(11):e78811. 25. Jenkinson M, Beckmann CF, Behrens TE, 14. Smets EM, Garssen B, Bonke B, De Haes Woolrich MW, Smith SM. FSL. Neuroimage JC. The Multidimensional Fatigue Inventory 2012;62(2):782–790.

526 radiology.rsna.org n Radiology: Volume 274: Number 2—February 2015