Neuroanatomic Connectivity of the Human Ascending Arousal ...ORIGINAL ARTICLE Neuroanatomic...

ORIGINAL ARTICLE

Neuroanatomic Connectivity of the Human Ascending ArousalSystem Critical to Consciousness and Its Disorders

Brian L. Edlow, MD, Emi Takahashi, PhD, Ona Wu, PhD, Thomas Benner, PhD, Guangping Dai, PhD,Lihong Bu, MD, PHD, Patricia Ellen Grant, MD, David M. Greer, MD, MA,

Steven M. Greenberg, MD, PhD, Hannah C. Kinney, MD, and Rebecca D. Folkerth, MD

AbstractThe ascending reticular activating system (ARAS) mediates arousal,

an essential component of human consciousness. Lesions of the ARAScause coma, the most severe disorder of consciousness. Because ofcurrent methodological limitations, including of postmortem tissueanalysis, the neuroanatomic connectivity of the human ARAS is poorlyunderstood. We applied the advanced imaging technique of highangular resolution diffusion imaging (HARDI) to elucidate the struc-tural connectivity of the ARAS in 3 adult human brains, 2 of whichwere imaged postmortem. High angular resolution diffusion imagingtractography identified the ARAS connectivity previously described inanimals and also revealed novel human pathways connecting thebrainstem to the thalamus, the hypothalamus, and the basal forebrain.Each pathway contained different distributions of fiber tracts fromknown neurotransmitter-specific ARAS nuclei in the brainstem. Thehistologically guided tractography findings reported here provide ini-tial evidence for human-specific pathways of the ARAS. The uniquecomposition of neurotransmitter-specific fiber tracts within each ARASpathway suggests structural specializations that subserve the differentfunctional characteristics of human arousal. This ARAS connectivityanalysis provides proof of principle that HARDI tractography may

affect the study of human consciousness and its disorders, including inneuropathologic studies of patients dying in coma and the persistentvegetative state.

Key Words: Arousal, Ascending reticular activating system (ARAS),Brainstem, Consciousness, High angular resolution diffusion imaging(HARDI), Neuroanatomy, Tractography.

INTRODUCTIONHuman consciousness consists of 2 critical components:

arousal and awareness (1, 2). Arousal pathways originating inthe brainstem activate awareness networks in the cerebralcortex via synapses in the thalamus and basal forebrain (3Y5)or, alternatively, via direct innervation of the cortex itself(4, 5). Without arousal, awareness is not possible, as evi-denced by comatose patients with brainstem lesions but ananatomically intact cerebral cortex (6, 7). The physiologicaland neuroanatomic basis of arousal in the brainstem has his-torically been conceptualized as the ascending reticular acti-vating system (ARAS), an idea introduced by Moruzzi andMagoun in 1949 (8). In classic studies in cats, electrical stim-ulation of the dorsal midbrain produced widespread bihemi-spheric activation of the cerebral cortex, as demonstrated byelectroencephalography (EEG). The pathways of the ARASwere initially thought to originate solely in the central core ofthe upper brainstem called the reticular formation becauseof its netlike histologic appearance (5). In this original modelof the ascending arousal system, neural projections from thereticular formation (e.g. the cuneiform/subcuneiform nucleusin the midbrain and pontis oralis in the rostral pons) werebelieved to activate the cerebral cortex via excitatory (gluta-matergic) relays in the thalamus.

It is now recognized that the ARAS is composed of acomplex and diffuse network of neurons projecting from mul-tiple brainstem source nuclei (within and adjacent to the classicreticular core) to the cortex via thalamic (9) and extrathalamicpathways (1, 2, 4, 5, 10). These pathways are typically called‘‘neurotransmitter specific’’ and include serotonergic fibers fromthe raphe subnuclei of the rostral pons and midbrain (11), nor-adrenergic fibers from the locus coeruleus of the rostral pons(12), dopaminergic fibers from the ventral tegmental area of thecaudal midbrain (13), cholinergic fibers from the pedunculo-pontine nucleus and laterodorsal tegmental nucleus of thecaudal midbrain and rostral pons (3), and glutamatergic fibersfrom the parabrachial complex in the rostral pons (14).

531J Neuropathol Exp Neurol � Volume 71, Number 6, June 2012

J Neuropathol Exp NeurolCopyright � 2012 by the American Association of Neuropathologists, Inc.

Vol. 71, No. 6June 2012

pp. 531Y546

Departments of Neurology (BLE) and Pathology (RDF), Brigham andWomen’s Hospital; Division of Newborn Medicine (ET, PEG), Fetal-Neonatal Neuroimaging and Developmental Science Center (ET, PEG),F.M. Kirby Neurobiology Center (LB), and Departments of Radiology(PEG) and Pathology (HCK, RDF), Children’s Hospital Boston, HarvardMedical School, Boston; Department of Neurology, J. Philip KistlerStroke Research Center (BLE, DMG, SMG) and Department of Radiol-ogy (OW), Massachusetts General Hospital, Harvard Medical School,Boston; Athinoula A. Martinos Center for Biomedical Imaging (BLE,ET, OW, TB, GD, PEG), Massachusetts General Hospital, Charlestown,Massachusetts; and Department of Neurology (DMG), Yale-NewHaven Hospital, Yale University School of Medicine, New Haven,Connecticut.

Send correspondence and reprint requests to: Brian L. Edlow, MD, Depart-ment of Neurology, Brigham and Women’s Hospital, 75 Francis St,Boston, MA 02115; E-mail: [email protected]

This work was supported by grants from the National Institutes of Health(R25 NS065743 [B.L.E.], R01 HD20991 [H.C.K.], R21 HD069001 [E.T.],and P41 RR14075 [Athinoula A. Martinos Center for Biomedical Imaging])and by the Center for Integration of Medicine & Innovative Technology, andby the Neuropathology Division, Department of Pathology, Brigham andWomen’s Hospital, Boston, MA. This work also involved the use of in-strumentation supported by the National Center for Research Resources(1S10RR016811-01).

Supplemental digital content is available for this article. Direct URL citationsappear in the printed text and are provided in the HTML and PDF versionsof this article on the journal’s Web site (www.jneuropath.com).

Copyright © 2012 by the American Association of Neuropathologists, Inc. Unauthorized reproduction of this article is prohibited.

by guest on October 8, 2016

http://jnen.oxfordjournals.org/D

ownloaded from

http://jnen.oxfordjournals.org/

Arousal is further mediated by ARAS connectivity with thehypothalamus, which participates in the regulation of auto-nomic function (15) and circadian sleep-wake cycles (16), andwith the basal forebrain, which participates in cortical acti-vation and autonomic integration (5, 14). Thus, in this report,‘‘ARAS’’ refers to the reticular core and extended sourcenuclei in the brainstem that mediate arousal, as well as theirrostral projections to the hypothalamus, the thalamus, thebasal forebrain, and the cortex. Of note, these extended sourcenuclei are the so-called neurotransmitter-specific nuclei. Inthis modern model of the ARAS, the thalamus is not simply arelay center; rather, it integrates and modulates the inter-actions between brainstem arousal networks and corticalawareness networks (17).

The elucidation of the neuroanatomic basis of thehuman ARAS is essential for determining the structural fea-tures of human arousal and for treating disorders of conscious-ness, such as traumatic or stroke-related coma. In addition,postmortem analyses of ARAS connections and their disruptionare of major importance for neuropathologists in the elucida-tion of the neuroanatomic substrate of coma (7), the persistentvegetative state (18), and the minimally conscious state (19)directly in the human brain. Yet, current neuroanatomic mod-els of the human ARAS are based largely on extrapolationsfrom animal studies, which may not be directly relevant tohumans given the unique features of human consciousness.Indeed, it is unknown which pathways in the human ARASare evolutionarily conserved and which pathways may haveformed new connections during evolutionary development ofthe human arousal system, as suggested by prior studiesshowing interspecies differences in brainstem connectivity(20). The critical methodological barrier preventing a detailedconnectivity analysis of the human ARAS at autopsy is thelimited feasibility of histologic tract tracing using postmortemdye injections. This limitation is attributable to excessivelylong diffusion times (i.e. months) and inability of tracers todiffuse long distances along myelinated axons (21, 22). More-over, conventional magnetic resonance imaging (MRI) of thebrainstem does not provide sufficient resolution to identify thesmall (i.e. millimeters), discrete components of the ARAS.Although functional neuroimaging studies in humans haverevealed activation in the brainstem and thalamus duringarousal (23), these studies do not provide information aboutneuroanatomic connectivity between different network nodes.Even tractography reconstructions of diffusion tensor MRIdata lack the angular resolution needed to identify the cross-ing nerve fibers (24) that are a prominent structural feature ofthe ARAS (5).

Recently, a more sophisticated magnetic resonance tech-nique, high angular resolution diffusion imaging (HARDI) trac-tography (25), has advanced the study of complex neuralnetworks by enhancing crossing fiber detection (26). Similar todiffusion tensor tractography, HARDI tractography is based onthe principle that the neuroanatomic trajectory of axon bundlescan be delineated by measuring the directionality of water dif-fusion along these axons (27, 28). The major methodologicaladvantage of HARDI tractography over diffusion tensor trac-tography is the ability to resolve multiple axonal bundles tra-versing in different directions within the same volume of

space, or voxel (26). High angular resolution diffusion imagingtractography may, therefore, provide greater spatial and angularresolution for connectivity analyses in the adult human brainthan currently available imaging or tissue labeling methods.We hypothesized that HARDI tractography elucidates thecomplex neuroanatomic connectivity of the human ARASin the brainstem, the hypothalamus, the thalamus, and thebasal forebrain. To test this hypothesis, we performed ARASconnectivity analyses in 2 postmortem human brain speci-mens, including one in which extensive histologic correlationanalyses were performed and in 1 living human subject. Weanalyzed the components of the ARAS that originate in theclassic reticular core and the extended neurotransmitter-specific nuclei. We refer here to these latter nuclei relative totheir known key neurotransmitter according to convention,although we did not perform neurotransmitter-specific immu-nocytochemical analysis of the postmortem brains. Also ofnote, we define ‘‘connectivity’’ between neuroanatomic regionsby the presence of a bundle of fiber tracts, the ends of whichterminate within each respective region. This definition ofstructural connectivity, as delineated by HARDI tractography,does not prove synaptic connectivity, which will requirefuture correlative structural-functional studies of the ARAS.

MATERIALS AND METHODS

Human Subjects: Clinical Information andAutopsy Findings

The brains of 3 adults without neurologic disease wereanalyzed with HARDI tractography: 2 were brains fromautopsy (Cases 1 and 2) and 1 was in a living subject (Case 3).Autopsies were performed in Cases 1 and 2 with consent ofthe family and included permission for research under ourinstitutional review boardYapproved protocol. For Case 3, anin vivo HARDI scan was performed with consent from thesubject under a separate institutional review boardYapprovedprotocol. The brain of Case 1 was used as the index case inwhich histologic data were correlated with the neuroanatomiclocalization of ARAS nuclei and the connectivity sites ofARAS fiber tracts, as identified by HARDI tractography.

Case 1 was a 53-year-old woman with a history ofbreast cancer (stage I) and a more recent diagnosis of high-grade, pleomorphic sarcoma involving the pelvis. She had nohistory of neurologic illness, brain metastases, or brain radi-ation. A normal neurologic examination was documentedduring her last hospitalization, 1 month before her death in along-term care facility. The death was attributed to systemiccomplications of cancer. At autopsy, the fresh brain weightwas normal (1,250 g; reference range, 1,200Y1,500 g). Theleptomeninges showed mild fibrosis. There was mild athero-sclerosis of the basilar, middle cerebral, and posterior cerebralarteries, but there were no occlusions or emboli. The cerebrumand cerebellum showed no macroscopic evidence of atrophy,hemorrhage, or infarcts. Standard sections in a survey of allbrain regions demonstrated mild arteriolosclerosis and agonalhypoxic-ischemic changes in the cerebellar cortex. Micro-scopic examination of the serial sections of the specimenimaged with HARDI tractography was unremarkable.

Edlow et al J Neuropathol Exp Neurol � Volume 71, Number 6, June 2012

� 2012 American Association of Neuropathologists, Inc.532




ownloaded from


Case 2 was a 49-year-old man with a history of hyper-tension, hyperlipidemia, diabetes, and T-cell prolymphocyticleukemia, and status post unrelated donor stem cell trans-plantation, who died of sepsis 48 days after transplantation.He had no history of neurologic disease or neurologic com-plications of his hematologic malignancy, and a normal resultin the neurologic examination was documented during hisfinal hospitalization. At autopsy, the fresh brain weight wasnormal (1,350 g). There was no evidence of atherosclerosis,occlusions, or emboli in the cerebral vasculature; the lep-tomeninges appeared normal. The cerebrum and cerebellumshowed no macroscopic evidence of atrophy, hemorrhage, orinfarcts. Standard sections in a survey of all brain regionswere stained with hematoxylin and eosin, Luxol fast blue formyelin, and Bodian silver for axons. Microscopic analysisshowed mild arteriolosclerosis and scattered reactive astro-cytes in gray and white matter. Two microscopic foci ofmyelin pallor on Luxol fast blue were noted: one in the rightperiventricular white matter and one in the cerebellar whitematter adjacent to the dentate nucleus, both immediatelyabutting the ependymal surface. These regions each measuredless than 0.3 cm and showed relative preservation of axons.There were no inflammatory infiltrates. Because of theseincidental findings, for which no clinical correlate could beidentified on review of all available medical records, anextensive analysis of the brainstem and diencephalic regionsthat were the focus of our tractography analysis was per-formed. No additional incidental lesions were identified in thepons, the midbrain, the hypothalamus, the thalamus, or thebasal forebrain on hematoxylin and eosin or Luxol fast bluestains. Case 3 was a 32-year-old man with no history ofmedical or neurologic disease.

Dissection and Imaging of Case 1 Brain SpecimenTo scan the ARAS with high spatial and angular reso-

lution, which is critical for tractography analyses of fiberpathways with complex branching patterns and high angles ofcurvature (24), a small horizontal-bore, high-field (4.7 T)Bruker Biospec MRI scanner was used in Case 1. To fit thespecimen into the small bore of the scanner, the cerebralhemispheres were dissected from the thalamus and basalforebrain, and the cerebellum was dissected from the brain-stem, such that the scanned specimen consisted of the pons,the midbrain, the hypothalamus, the thalamus, the basalforebrain, and the basal ganglia (Figure, Supplemental DigitalContent 1, http://links.lww.com/NEN/A333). The dimensionsof the dissected specimen were 7.0 cm (rostral-caudal axis) �4.0 cm (anterior-posterior axis at the level of the thalami) �5.9 cm (medial-lateral axis at the level of the thalami). Thedissection was performed 80 days after the patient’s death,and imaging data were acquired 103 days after death.

Immediately before scanning, the dissected brain specimenwas transferred from a 10% formaldehyde solution to a Fomblinsolution (perfluoropolyether, Ausimont USA, Inc., Thorofare,NJ), which reduces magnetic susceptibility artifacts that mayoccur during image acquisition (29). The HARDI sequence usedin this scan was a 3-dimensional (3D) diffusion-weighted spin-echo echo-planar imaging (DW-SE-EPI) sequence with 60diffusion-weighted measurements, corresponding to a cubic

lattice in Q-space of b = 4,057 s/mm2, using gradient strengthof 12.4 G/cm, duration ¯ = 13.4 milliseconds, and inter-temporal pulse offset $ = 25 milliseconds. Repetition time(TR) was 1,000 milliseconds, echo time (TE) was 72.5 mil-liseconds, the field of view was 7.2 � 7.8 � 8.2 cm, and theimaging matrix was 128 � 128 � 128 pixels, yielding aspatial resolution of 562 � 609 � 641 Km. One data set of 3DDW-SE-EPI data with b = 0 s/mm2 was also acquired tocalculate quantitative diffusion properties in each voxel. Totalimage acquisition time was 130 minutes.

Imaging of Case 2 Brain SpecimenThe brain specimen for Case 2 was scanned as a whole

brain on a 3-T Tim Trio MRI scanner (Siemens MedicalSolutions, Erlangen, Germany) using a 32-channel head coil.As in Case 1, the specimen for Case 2 was transferred from a10% formaldehyde solution to a Fomblin solution immedi-ately before imaging, which was performed 8 months 16 daysafter death. To increase diffusion sensitivity on the clinical3-T MRI scanner that was required to scan a whole brainspecimen, diffusion data were acquired using a 3D diffusion-weighted steady-state free-precession (DW-SSFP) sequence(30, 31), with a 3D FT readout using the following param-eters: TR = 27.8 milliseconds, TE = 22.9 milliseconds, flipangle = 60 degrees, bandwidth = 150 Hz/pixel, diffusion gra-dient duration = 18 milliseconds, diffusion gradient amplitude =3.2 G/cm, matrix size = 192 � 176 � 128 pixels, and field ofview = 19.2 � 17.6 � 12.8 cm, yielding a spatial resolutionof 1.0 � 1.0 � 1.0 mm. Four data sets of nonYdiffusion-weighted volumes (b = 0 s/mm2) and 44 data sets of diffusion-weighted volumes were acquired, resulting in a total scan timeof 5 hours 35 minutes. Of note, in a DW-SSFP sequence,diffusion weighting is not defined by a single global b value,as in the DW-SE-EPI sequence used in Case 1, because thediffusion signal in a DW-SSFP sequence cannot be readilydissociated from other imaging properties, such as the T1 re-laxation time, the T2 relaxation time, TR, and the flip angle(30, 32). Diffusion weighting in a DW-SSFP study is there-fore characterized by the diffusion gradient duration and am-plitude, as reported previously.

Imaging of Case 3Imaging was performed for Case 3 on a 3-T Tim Trio

MRI scanner (Siemens Medical Solutions) with a 32-channelhead coil. High angular resolution diffusion imaging datawere acquired using a twice-refocused SE-EPI sequence (33)with the following parameters: TR/TE = 12,750/130 milli-seconds, bandwidth = 1,395 Hz/pixel, matrix size = 128 �128, and field of view = 256 � 256 cm2, yielding an in-planespatial resolution of 2.0 � 2.0 mm. Seventy-four slices of2 mm slice thickness were acquired. Twenty data sets ofnonYdiffusion-weighted volumes (b = 0 s/mm2) and 120 datasets of diffusion-weighted volumes were acquired with aprespecified b value of 4,000 s/mm2, resulting in a total scantime of 29 minutes 45 seconds.

Histologic MethodsAt the completion of image acquisition, the dissected

brain specimen of Case 1, consisting of the pons, the midbrain,

J Neuropathol Exp Neurol � Volume 71, Number 6, June 2012 Human Ascending Arousal System Connectivity

� 2012 American Association of Neuropathologists, Inc. 533




ownloaded from


the hypothalamus, the thalamus, the basal forebrain, and thebasal ganglia en bloc (Figure, Supplemental Digital Content 1,http://links.lww.com/NEN/A333) was divided into 7 blocks,approximately 0.5 to 1.5 cm in thickness, which were thenembedded in paraffin. Serial transverse sections of the ponsand the midbrain and coronal sections of the hypothalamus,the thalamus, and the basal forebrain were prepared at 10-Kmthickness with a microtome (Leica RM2255; Leica Micro-systems, Buffalo Grove, IL). Every 50th section was stainedwith hematoxylin and eosin for nuclear boundaries and wascounterstained with the myelin stain, Luxol fast blue, for fibertract boundaries. Each examined section was separated by500 Km, for a total of 74 sections from mid pons throughthe entire hypothalamus, thalamus, basal forebrain, and basalganglia.

Regions of Interest for ARAS TractographyAscending reticular activating system fiber tracts were

identified using regions of interest (ROIs) that were manuallytraced on the diffusion images using TrackVis (version 5.1), aninteractive image analysis software program that is availableto the scientific community without charge (34). For Case 1,each diffusion image was compared with its correspondinghistologic section to ensure that radiologic ROIs shared thesame borders, size, and contours as the histologic ROIs. His-tologic ROIs were delineated by visual inspection with a lightmicroscope and were confirmed by standard atlases of humanneuroanatomy (35Y37). Regions of interest were identified for allof the key ARAS source nuclei implicated in arousal: cunei-

form/subcuneiform nucleus, pontis oralis, median and dorsalraphe, locus coeruleus, pedunculopontine nucleus, parabrachialcomplex (combined medial and lateral parabrachial nuclei), andventral tegmental area (Figs. 1A, B; Figure, SupplementalDigital Content 2, http://links.lww.com/NEN/A336 and Figure,Supplement Digital Content 3, http://links.lww.com/NEN/A337).Regions of interest were also traced using histologic guidancefor the thalamic nuclei implicated in the modulation, or gating,of arousal: the reticular nucleus, the central lateral nucleus, andthe centromedian/parafascicular nuclear complex (Figure, Sup-plemental Digital Content 2, http://links.lww.com/NEN/A336).Each brainstem and thalamic ROI served as a seed point fromwhich fiber tracts were generated. Because the brainstem ROIsare known to change in shape, size, and contour along the ros-trocaudal axis, histologic guidance of ROI tracing was per-formed for every axial diffusion image using its correspondinghistologic section. Similarly, each thalamic nucleus was tracedon the coronal diffusion images with direct guidance by thelocation of the stained nuclei on corresponding coronal tissuesections. For the intrathalamic connectivity analyses, a 2-ROI-tractography technique was used based on methods developedby Catani et al (38). Specifically, fiber tracts passing betweenthe reticular nucleus and central lateral nucleus were ‘‘virtuallydissected’’ from fiber tracts passing between the reticularnucleus and the centromedian/parafascicular nuclear complex.For Cases 2 and 3, brainstem and thalamic nuclei were tracedin accordance with the aforementioned neuroanatomic atlases.We also compared the neuroanatomic localization, contours,and boundaries of these ROIs in Cases 1, 2, and 3 to ensureconsistency in the tractography analyses.

FIGURE 1. Diffusion tractography of the human ascending reticular activating system, as defined by the cuneiform/subcuneiform(CSC) region of interest (ROI). (A, B) Transverse histologic (A) and radiologic (B) sections through the rostral midbrain at the levelof superior colliculi (SC) in Case 1. The image in B is a nonYdiffusion-weighted image with b = 0 s/mm2; the CSC ROI is labeled inred. Neuroanatomic landmarks: mamillary bodies (MB), cerebral peduncles (CP), red nuclei (RN), superior colliculi (SC), andperiaqueductal gray (PAG). (C, D) Ventral (C) and left lateral (D) views of CSC fiber tracts. Tracts are color-coded according totheir direction (red, medial-lateral; green, ventral-dorsal; blue, rostral-caudal). Bundles of CSC tracts are labeled as follows: VTTR,ventral tegmental tract, rostral; VTTC, ventral tegmental tract, caudal; DTTL, dorsal tegmental tract, lateral; DTTM, dorsal tegmentaltract, medial; and MFB, medial forebrain bundle.






ownloaded from


There are variations in nomenclature pertaining to thesource nuclei of the ARAS in standard neuroanatomic atlasesof the human brainstem. In this study, the neuroanatomiclocalization of the pedunculopontine nucleus was tracedaccording to the brainstem atlas of Paxinos and Huang (35),extending from the caudal midbrain to the caudal border ofthe red nuclei. The neuroanatomic localization of the para-brachial complex was traced according to the brainstem atlasof Olszewski and Baxter (36), extending from the mid pons tothe rostral pons.

Tract Construction and VisualizationDiffusion data were processed in Diffusion Toolkit

(version 6.0) (34) using a HARDI reconstruction in which theorientation distribution function of angular water diffusionis calculated for each voxel on a spherical harmonic basis(39). By calculating an orientation distribution function foreach voxel, the HARDI reconstruction allows for multiplediffusion directions within each voxel, as opposed to thesingle diffusion direction that is calculated in a diffusiontensor imaging reconstruction. Fiber tracking was performedin TrackVis (version 5.1) (34) using a streamline, determin-istic tractography method. Fiber tract propagation is based onthe coherence of directional water diffusion in adjacent voxels(40), as determined by the orientation distribution functionsgenerated by the HARDI reconstruction. When the directionof water diffusion is similar in adjacent voxels, the tractog-raphy model displays these voxels as being part of a coherentfiber tract. When directional diffusion in multiple voxels ispotentially coherent, the tractography algorithm consistentlypursues the water diffusion vector of least voxel-to-voxelcurvature, thereby reducing the possibility of identifyingspurious, inaccurate fiber tracts. In this experiment, fiber tractswere terminated when the angle between water diffusion vec-tors in adjacent voxels exceeded a prespecified threshold of60 degrees. Whereas tracts tend to terminate at sites of fibercrossing in diffusion tensor tractography, the improved angu-lar resolution provided by HARDI allows tracts to continuein multiple directions at sites of fiber crossing (26). Giventhat low measurements of directional water diffusion, orfractional anisotropy (FA), have been observed in the dorsalpons, the midbrain (41), and the thalamus (42) in humans, noFA threshold was applied to the tractography analyses. Forintrathalamic connectivity analyses, a length threshold wasapplied to eliminate the long fiber tracts connecting the tha-lamic nuclei to the brainstem. This length threshold wasnecessary to isolate the comparatively shorter intrathalamicfiber tracts.

To test whether the HARDI data were of sufficientquality for tractography reconstructions (i.e. connectivitymodeling), we calculated the signal-to-noise ratio for eachHARDI scan by dividing the mean signal within the whitematter of the internal capsule by the standard deviation of thesignal outside the specimen (i.e. noise) on the trace diffusion-weighted images and the nonYdiffusion-weighted images withb = 0 s/mm2 (b0). The diffusion-weighted images signal-to-noise ratio for Cases 1, 2, and 3 were 466, 186, and 137,respectively. The b0 signal-to-noise ratios were 49, 106, and308, respectivelyVvalues that are more than sufficient to

produce high-quality tractography results (25). For theHARDI scans in Cases 1 and 3, we also performed DiffusionToolkit diffusion tensor reconstructions for standard quanti-tative measurements of water diffusionVFA and the apparentdiffusion coefficient (ADC)Vbecause these measurementshave not previously been performed in ARAS fiber tracts. Forcomparison, FA and ADC were also measured within thecorticospinal tract, a structure in which water diffusion hasbeen well characterized in studies of ex vivo (43) and in vivo(44) human brains. The purpose of these comparative FA andADC analyses was to ascertain how water diffusion in thecomplex network of ARAS fiber tracts compared with waterdiffusion within the more homogeneous parallel fibers of thecorticospinal tract. Both FA and ADC values were not cal-culated for Case 2 because, as stated above, a single b valuecannot be precisely calculated for the DW-SSFP sequencethat was used in Case 2. As a result, whereas the DW-SSFPsequence provides well-validated measurements of the direc-tionality of water diffusion for tractography modeling in thehuman brain (31, 45), the complex signal evolution in theDW-SSFP sequence (32) precludes standard quantitative dif-fusion measurements (31).

To optimize the specificity of ARAS tract identification,we rigorously eliminated non-ARAS fiber tracts from theARAS tractography analysis. Non-ARAS ROIs that were inclose neuroanatomic proximity to ARAS ROIs were firstidentified and traced on the diffusion images for Case 1 usinghistologic guidance, and then all fiber tracts passing throughthese non-ARAS ROIs were excluded from subsequentARAS tractography analyses. We eliminated all fiber tractspassing through the superior cerebellar peduncle (both thebrachium and the decussation), middle cerebellar peduncle,cranial nerve III, cerebral peduncle (including the fronto-pontine fibers), medial lemniscus, ventral trigeminothalamictract (which runs dorsal to the medial lemniscus in the rostralpons), and the medial longitudinal fasciculus (which ascendsat the ventromedial margin of the periventricular and peri-aqueductal gray). Once these non-ARAS fiber tracts wereeliminated, ARAS connectivity was determined by analyzingthe trajectories and termination sites of the ARAS fiber tractswithin the brainstem, the hypothalamus, the thalamus, thebasal forebrain, and the basal ganglia. For Cases 2 and 3, non-ARAS pathways were eliminated by tracing non-ARAS ROIsusing guidance from neuroanatomic atlases. Because theHARDI scans for Cases 2 and 3 were performed on wholebrains, we also eliminated fiber tracts from the fornix, astructure that was not present in the dissected specimen ofCase 1.

RESULTS

Connectivity of the Cuneiform/SubcuneiformNucleus and Pontis Oralis

High angular resolution diffusion imaging tractographyrevealed a bilateral fiber bundle connecting the cuneiform/subcuneiform nucleus in the rostral midbrain to the thala-mus, hypothalamus, and basal forebrain (Table 1; Fig. 1).The originating source of this fiber bundle also includedthe pontis oralis of the rostral pons (Fig. 2). In the rostral






ownloaded from


midbrain, the bilateral fiber bundle bifurcated into ventral anddorsal bundles (Figs. 1Y3A). Here we suggest the labelsventral tegmental tract (VTT) and dorsal tegmental tract(DTT) for these 2 large divergent bundles in the human brain,names with historical precedent in experimental animals(46Y48). The human VTT bifurcated yet again into 2 addi-tional fiber bundles: one is a distinct caudal pathway con-necting with the hypothalamus, zona incerta, Forel fields, basal

forebrain, and globus pallidus; and the other, a rostral pathwayconnecting with the paraventricular (midline) region of thethalamus (Table 1; Figs. 3A and 4). We labeled the humanhypothalamic VTT pathway as the VTT caudal (VTTC), andthe thalamic VTT pathway as the VTT rostral (VTTR).

The human DTT bifurcated into 2 additional fiber bun-dles that each projected to a different distribution of thalamicnuclei. One subdivision, which we labeled the DTT lateral

TABLE 1. Neuroanatomic Connectivity of the Human Ascending Reticular Activating System

Source Nucleus

Thalamus

Hypothalamus Basal Forebrain FF/ZI Globus PallidusIL Ret PV Pul LGN

CSC DTTM DTTL DTTM VTTR DTTL DTTL VTTC VTTC DTTL VTTC VTTC

PO DTTM DTTL DTTM VTTR DTTL DTTL VTTC VTTC VTTC VTTC

PPN DTTM DTTL DTTM VTTR DTTL DTTL VTTC VTTC DTTL VTTC VTTC

PBC DTTM DTTL DTTM VTTR DTTL DTTL VTTC VTTC DTTL VTTC VTTC

DR DTTM DTTL VTTR V V VTTC VTTC VTTC V

MR V V V V V VTTC VTTC VTTC V

LC DTTM DTTL VTTR V V VTTC VTTC VTTC V

VTA V V V V V VTTC VTTC VTTC VTTC

The pathways that connect each region are the dorsal tegmental tract, medial (DTTM); dorsal tegmental tract, lateral (DTTL); ventral tegmental tract, caudal (VTTC); and ventraltegmental tract, rostral (VTTR).

‘‘V,’’ no connectivity.CSC, cuneiform/subcuneiform nucleus; DR, dorsal raphe; FF/ZI, Forel fields/zona incerta; IL, intralaminar nuclei of the thalamus; LC, locus coeruleus; LGN, lateral geniculate

nucleus of the thalamus; MR, median raphe; PV, paraventricular region of the thalamus; PBC, parabrachial complex; PPN, pedunculopontine nucleus; PO, pontis oralis; Pul, pulvinar;Ret, reticular nucleus of the thalamus; and VTA, ventral tegmental area.

FIGURE 2. Diffusion tractography of the cuneiform/subcuneiform nucleus (CSC) and pontis oralis (PO) pathways. (A, B) Left lateral(A) and zoomed left lateral (B) views of the CSC and PO regions of interest, and the fiber tracts that pass through these regions inCase 1. Fiber tracts are color-coded by the source nucleus of origin (red, CSC; light blue, PO). Tracts in A are superimposed on anaxial nonYdiffusion-weighted image with b = 0 s/mm2 at the level of the rostral pons. Bundles of CSC and PO fiber tracts are labeledas follows: VTTR, ventral tegmental tract, rostral; VTTC, ventral tegmental tract, caudal; DTTL, dorsal tegmental tract, lateral; DTTM,dorsal tegmental tract, medial; and MFB, medial forebrain bundle. Of note, red fiber tracts generated by the CSC region of interestextend both rostrally and caudally, but the direction of electrical signaling within each fiber tract (i.e. ascending or descending)cannot be determined.






ownloaded from


(DTTL), projected laterally to the thalamic reticular nucleusand, to a lesser extent, the basal forebrain, pulvinar, and lateralgeniculate nucleus of the thalamus (Table 1; Figs. 1C, D, 2,3A, and 5). The second subdivision of the DTT, which welabeled DTT medial (DTTM), projected medially to the thalamicintralaminar nuclei (central lateral nucleus and centromedian/parafascicular nuclear complex). In addition, several DTTM

fibers projected beyond the intralaminar nuclei along a rostraland lateral course to the reticular nucleus (Figs. 3A and 5). Ofnote, examination of the cuneiform/subcuneiform fiber tractsrevealed both short and long rostral projections and fiberscrossing to the contralateral side of the brainstem (Figure, Sup-plemental Digital Content 4, http://links.lww.com/NEN/A338),well-recognized features of the brainstem reticular formation(48). In Cases 2 and 3, tractography analyses of ARAS brain-stem ROIs revealed VTTR, VTTC, DTTL, and DTTM pathwayswhose trajectories and sites of connectivity were consistent withthe findings in Case 1 (Fig. 5; Figure, Supplemental DigitalContent 5, http://links.lww.com/NEN/A339).

Comparative FA and ADC analyses in the cuneiform/subcuneiform, pontis oralis, and corticospinal fiber tractsdemonstrated that FA values were lower and ADC valueswere higher in the cuneiform/subcuneiform and pontis oralisfiber tracts than in the corticospinal fiber tracts for both Case 1and Case 3 (Table 2). Both FA and ADC measurements in thecorticospinal fiber tracts in Case 1 were consistent with thosereported for ex vivo human brains (43). Of note, the lowerADC values observed in Case 1 compared to Case 3 areexpected because ADC values have been demonstrated to belower in the postmortem human brain than in the livinghuman brain (43). Interestingly, FA values were similar inCases 1 and 3, consistent with prior studies showing that tis-sue death and fixation do not affect FA as much as ADC (43).Whereas there are few prior studies of FA or ADC in livinghuman subjects imaged at high b values, our corticospinaltract FA and ADC measurements are generally consistentwith the results of Yoshiura et al (49), who measured FA andADC in the internal capsule using the same b value of 4,000 s/mm2 that was used in Case 3. Also of note, the in vivo whitematter ADC values of approximately 400 to 500 � 10-6 mm2/sacquired with a b value of 4,000 s/mm2 in this study and inPierpaoli et al (50) are significantly lower than the ADC val-ues of approximately 700 � 10j6 mm2/s that have beenreported in prior in vivo studies of human white matter usingb values ranging from 700 to 1,000 s/mm2 (49, 50). Thisdecline in ADC is explained by the nonmonoexponential rela-tionship between b value and ADC at high b values (wherehigh is typically considered as b 9 2,000 s/mm2 for in vivohuman studies) (49, 51).

Connectivity of Monoaminergic-, Cholinergic-,and Glutamatergic-Specific ARAS Nuclei

Subdivisions of the VTT and DTT were further dem-onstrated to carry fibers of the known monoaminergic-,cholinergic-, and glutamatergic-related source nuclei inthe brainstem. Each of these source nuclei had specificpatterns of connectivity with the thalamus (Figs. 3A,6, and 7; Figure, Supplemental Digital Content 6,http://links.lww.com/NEN/A341; Figure, Supplemental DigitalContent 7, http://links.lww.com/NEN/A342; Figure, SupplementalDigital Content 8, http://links.lww.com/NEN/A343; Figure, Sup-plemental Digital Content 9, http://links.lww.com/NEN/A344; Figure,Supplemental Digital Content 10, http://links.lww.com/NEN/A345),hypothalamus (Figs. 4, 6, and 7; Figure, Supplemental DigitalContent 6, http://links.lww.com/NEN/A341; Figure, Supplemental

FIGURE 3. Neuroanatomic connectivity of brainstem arousalpathways and intrathalamic gating pathways. (A) Dorsal viewof fiber tracts originating in the pedunculopontine nucleus(purple) and parabrachial complex (yellow) connecting withthe centromedian/parafascicular (CEM/Pf) and central lateral(CL) nuclei via the dorsal tegmental tract, medial (DTTM) inCase 1. Pedunculopontine and parabrachial fibers also connectwith the reticular nuclei (Ret) via the dorsal tegmental tract,lateral (DTTL), and via an extension of DTTM (arrowheads).Fiber tracts from the dorsal raphe (turquoise), locus coeruleus(dark blue), pedunculopontine nucleus (purple), and para-brachial complex (yellow) connect with the paraventricular,midline nuclei of the thalamus via the ventral tegmental tract,rostral (VTTR). (B) Dorsal view of intrathalamic fiber tractsconnecting Ret with CL (red) and Ret with CEM/Pf (pink). Fibertracts in A and B are superimposed on an axial nonYdiffusion-weighted image with b = 0 s/mm2 (b0) at the level of the rostralmidbrain and a coronal b0 image at the level of the mid tha-lamus. Axial and coronal images in A are semitransparent, sothat the trajectories of VTTR fiber tracts are visualized coursingrostrally and dorsally to the paraventricular region of the tha-lamus. Neuroanatomic landmarks: red nuclei (RN), superiorcolliculi (SC), and periaqueductal gray (PAG).






ownloaded from


Digital Content 7, http://links.lww.com/NEN/A342; Figure, Supple-mental Digital Content 8, http://links.lww.com/NEN/A343; Figure,Supplemental Digital Content 9, http://links.lww.com/NEN/A344; Fig-ure, Supplemental Digital Content 10, http://links.lww.com/NEN/A345),and basal forebrain (Table 1; Figure, Supplemental DigitalContent 7, http://links.lww.com/NEN/A342). Of note, diffusethalamic projections that are known to arise from serotonergicneurons in the raphe and noradrenergic neurons in the locuscoeruleus were not identified with ex vivo HARDI tractography inCase 1 but were identified with ex vivo HARDI tractography in

Case 2 and in vivo HARDI tractography in Case 3 (Figure, Sup-plemental Digital Content 11, http://links.lww.com/NEN/A346).

Intrathalamic ConnectivityWe delineated connections between the thalamic nuclei

implicated in gating of the inputs of the ascending arousal systemto the cerebral cortex (17, 52). The reticular nucleus connectedwith both the central lateral nucleus and the centromedian/parafascicular nuclear complex via dense bundles of fiberswith direct, linear trajectories (Fig. 3B). These intrathalamic

FIGURE 4. Ascending reticular activating system (ARAS) connectivity with the hypothalamus. (A, B) Ventral (A) and rostral (B)views of ARAS connectivity with the hypothalamus in Case 1. (C) Ventral view from A with display of tract end points. All fibertracts and end points are color-coded according to their source nucleus of origin: turquoise, dorsal raphe; green, median raphe;dark blue, locus coeruleus; purple, pedunculopontine nucleus; yellow, parabrachial complex; and pink, ventral tegmental area.End points for all source-nucleus tracts terminate in the hypothalamus but only dorsal raphe, locus coeruleus, pedunculopontine,and parabrachial complex pathways connect with the anterior region of the hypothalamus (arrowheads, arrows, and asterisks inAYD). (D) High-zoom view of tract end points in the left anterior hypothalamus demonstrates termination sites of locuscoeruleus, pedunculopontine, parabrachial complex, and dorsal raphe pathways, some of which overlap within the same voxel(dimensions 562 � 609 � 641 Km). All fiber tracts and end points are superimposed on an axial nonYdiffusion-weighted imagewith b = 0 s/mm2 (b0) at the level of the red nuclei and a coronal b0 image at the level of the mid thalamus (inset). The axial andcoronal b0 images in A are semitransparent so that the trajectories of fiber tracts are seen as they cross each respective plane.Bundles of fiber tracts are labeled as follows: VTTR, ventral tegmental tract, rostral; VTTC, ventral tegmental tract, caudal. Neu-roanatomic landmarks: 3V, third ventricle; CP, cerebral peduncles; and RN, red nuclei.






ownloaded from


fiber bundles were interspersed with the DTTM fibers travelingthrough the intralaminar nuclei to the reticular nucleus (Figure,Supplemental Digital Content 12, http://links.lww.com/NEN/A347).Also of note, a group of fiber tracts was observed to connect the

reticular nucleus with Forel fields and zona incerta by passingthrough the central lateral and centromedian/parafascicular nuclei(Figs. 3B and 6). The termination points of these fiber tractswithin Forel fields and zona incerta were near the termination

FIGURE 5. Connectivity of the cuneiform/subcuneiform nucleus (CSC) and pontis oralis (PO) in Cases 1, 2, and 3. (AYC) A dorsalview of left-sided CSC fiber tracts (red) and PO fiber tracts (blue) is shown for all 3 high angular resolution diffusion imaging(HARDI) scans (A, Case 1; B, Case 2; C, Case 3). For each scan, a summary of key HARDI sequence parameters is provided below thefiber tracts; the 3 planes (sagittal, coronal, and axial) onto which the fiber tracts are superimposed are shown as nonYdiffusion-weighted images with b = 0 s/mm2. The 2 divergent branches of the dorsal tegmental tract (DTT)VDTT lateral (DTTL) and DTTmedial (DTTM)Vare seen in each case. In addition, each HARDI scan demonstrates that DTTM connects with and passes throughthe central lateral nucleus (red, CL) and the centromedian/parafascicular complex (pink, CEM/Pf) of the thalamus, and a bundle ofDTTM fiber tracts projects rostrally and laterally to the reticular nucleus of the thalamus (purple, Ret) (white arrows, A, B, C). Ofnote, the Ret, CL, and CEM/Pf thalamic nuclei are shown in only 1 coronal slice in B and C so that the fiber tracts can be optimallyvisualized, whereas these nuclei are shown as 3-dimensional structures in A. Neuroanatomic landmarks: CC, splenium of the corpuscallosum; IC, inferior colliculus; P, pineal gland; and Thal, thalamus.

TABLE 2. Fractional Anisotropy and Apparent Diffusion Coefficient Measurements for Cuneiform/Subcuneiform, Pontis Oralis, andCorticospinal Fiber Tracts in Cases 1 and 3

Case No.

Cuneiform/Subcuneiform Fiber Tracts Pontis Oralis Fiber Tracts Corticospinal Fiber Tracts

FA ADC (�10j6 mm2/s) FA ADC (�10j6 mm2/s) FA ADC (�10j6 mm2/s)

Case 1 (ex vivo) 0.25 T 0.14 160.2 T 35.8 0.27 T 0.09 156.4 T 31.7 0.43 T 0.12 110.4 T 32.9

Case 3 (in vivo) 0.25 T 0.14 512.6 T 200.2 0.29 T 0.14 482.6 T 190.4 0.45 T 0.18 438.4 T 176.8

Both FA and ADC values were measured within all fiber tracts projecting from each region of interest. Data are provided as mean T SD. In the corticospinal fiber tract analysis,tracts were generated using bilateral regions of interest traced on the cerebral peduncles at the level of the red nuclei in the midbrain. To increase the specificity of corticospinal tractanalysis, only fiber tracts that also passed through a region of interest traced in the posterior limb of the internal capsule at the level of the mid thalamus were included in the analysis. FAvalues were lower and ADC values were higher in the CSC and PO fiber tracts compared with the corticospinal fiber tracts, a finding that may be explained by the complex branching offiber tracts in the CSC and PO pathways, compared with the parallel composition of fiber tracts in the corticospinal pathway.

ADC, apparent diffusion coefficient; CSC, cuneiform/subcuneiform; FA, fractional anisotropy; PO, pontis oralis.






ownloaded from


points of brainstem source nuclei pathways that connectedwith these same regions (Fig. 6).

DISCUSSIONThis study provides proof of principle that HARDI trac-

tography is a potent tool for dissecting the complex neuro-

anatomic substrate of the human ARAS. Virtual dissection ofARAS pathways was demonstrated in the human brain, bothin vivo and ex vivo, suggesting that HARDI tractography maybe used in living patients and in postmortem neuropathologicanalyses to study the neuroanatomic basis of consciousness andits disorders. We found that the human ARAS shares several

FIGURE 6. Connectivity of intrathalamic gating pathways and brainstem arousal pathways projecting to the thalamus, hypothal-amus, Forel fields (FF), and zona incerta (ZI). (A) Dorsal view of intrathalamic fiber tracts that connect the reticular nucleus (Ret)with centromedian/parafascicular complex (CEM/Pf, pink fiber tracts) and that connect Ret with the central lateral nucleus (CL, redfiber tracts) in Case 1. Many red Ret-CL fiber tracts extend caudally into CEM/Pf, FF, and ZI, and pink Ret-CEM/Pf tracts also extendinto FF and ZI. Brainstem arousal pathways are displayed as fiber tract end points, color-coded by the source nucleus from whichthe tracts are generated: pedunculopontine nucleus (purple), parabrachial complex (yellow), dorsal raphe (turquoise), locuscoeruleus (dark blue), median raphe (green), and ventral tegmental area (pink). (B) High-zoomed dorsal view from within the leftCEM/Pf, which is semitransparent to allow for visualization of intrathalamic fiber tracts and pedunculopontine tract end points(arrow) within this nuclear complex. (C) Zoomed dorsal view of right side of image in A shows that pedunculopontine fiber tractsterminate in FF and ZI (purple end points) near fiber tracts that connect these regions to Ret via pathways that pass through CEM/Pfand CL (arrowheads, A, C). Fiber tracts in A to C are superimposed on an axial nonYdiffusion-weighted image with b = 0 s/mm2

(b0) at the level of the rostral midbrain and a coronal b0 image at the mid thalamus. Neuroanatomic landmarks: third ventricle (3V),red nuclei (RN), superior colliculi (SC), and periaqueductal gray (PAG).






ownloaded from


important features with that of experimental animals: 1) ARASbrainstem source nuclei connect primarily with the intralaminar,paraventricular, and reticular nuclei of the thalamus; 2) theknown monoaminergic-, cholinergic-, and glutamatergic-relatednuclei in the upper brainstem all connect with the hypothalamus;3) the basal forebrain is connected with multiple brainstemsource nuclei of the ARAS; and 4) the reticular, central lateral,and centromedian/parafascicular thalamic nuclei connect witheach other, providing a structural basis for human intrinsic tha-lamic networks that modulate ARAS activation of thalamocort-

ical networks. The short and long rostral projections of thecuneiform/subcuneiform tracts and the fibers crossing to thecontralateral side of the brainstem in our 3 cases are also well-recognized features of the brainstem reticular formation (48) andare consistent with the recognized complexity of the ARASnetwork (5). Yet, we also defined new human pathways, theVTT and DTT and their subdivisions, which are distinct in theirtrajectory, connectivity, and source nuclei origins from the VTTand DTT pathways described in animals with tract tracingmethods (46Y48). In the following discussion, we highlight our

FIGURE 7. The neuroanatomic connectivity of the serotonergic-related dorsal raphe (DR). (A) Left lateral oblique view of DR fibertracts (turquoise) in Case 1 superimposed on an axial nonYdiffusion-weighted image with b = 0 s/mm2 (b0) at the level of themidpons and a coronal b0 image at the level of the mid thalamus. The images are semitransparent so that the trajectories of fibertracts are seen as they cross the coronal plane. (B) Left lateral view superimposed on an axial b0 image at the level of the mid ponsand a sagittal b0 image at the midline of the specimen. (C)Dorsal view superimposed on an axial b0 image at the level of the inferiorcolliculi (caudal midbrain) and a coronal b0 image at the mid thalamus. Fiber pathways from DR are seen connecting with theinferior colliculi bilaterally (arrows). Bundles of fiber tracts are labeled as follows: VTTR, ventral tegmental tract, rostral; VTTC, ventraltegmental tract, caudal. Neuroanatomic landmarks: red nuclei (RN), thalamus (Thal), and pineal gland (P).






ownloaded from


findings in the context of the current understanding of ARASneuroanatomy, with emphasis on how our data expand on pre-vious findings in animals and humans. We begin with a con-sideration of key methodological issues and limitations ofHARDI tractography for consideration in the interpretation ofour data.

Methodological Considerations and Limitationsin HARDI Tractography Analyses

It is important to emphasize that HARDI tractographyprovides an inferential model, and not a direct measurement,of white matter connectivity in the human brain. In HARDIstudies, the accuracy of the neuroanatomic connectivity modeldepends on several methodological factors that must be con-sidered in both the data acquisition and postprocessing stagesof the experiment. For example, the sensitivity of HARDItractography for detecting fiber tracts depends primarily onthe spatial and angular resolution of the water diffusionmeasurements that are performed during the scan (dataacquisition), which provide the basis for reconstructing amodel of white matter connectivity. The specificity of HARDItractography data is related to the ability to eliminate fibertracts from nonrelevant neuroanatomic pathways (data post-processing). In this study, we optimized the sensitivity ofARAS tract identification by acquiring HARDI data withextremely high spatial and angular resolution, particularly inthe index case (Case 1), as evidenced by the high number ofdiffusion gradients (n = 60) (53, 54), the strength of the gra-dients (12.4 G/cm), which allows for high b values with rel-atively short gradient duration times (55), and the high fieldstrength of the magnet (4.7 T), which allows for high in-planeresolution (56). Furthermore, the postmortem, ex vivo dataacquisition in Cases 1 and 2 further increase HARDI’s sen-sitivity for identifying ARAS pathways by allowing for along duration of scan time (2 hours 10 minutes for Case 1 and5 hours 35 minutes for Case 2), as well as the elimination ofmovement artifacts and susceptibility artifacts caused by theskull base and air-filled sinuses. In Case 3, we increased thesensitivity of in vivo HARDI data acquisition by using a highnumber of directional diffusion gradients (n = 120) and a highb value (4,000 s/mm2). The specificity of ARAS tract identi-fication was optimized during the postprocessing stage byrigorously excluding non-ARAS fiber tracts and by definingeach ARAS ROI as precisely as possible using histologicguidance, as discussed previously. However, it should benoted that the ARAS is composed of multisynaptic pathwaysthat ascend and descend throughout the brainstem, andbecause HARDI tractography does not provide functionaldata about the direction of electrical signaling within thesepathways, we cannot eliminate the possibility that some of theARAS pathways identified in this study contained descendingfiber tracts. On the other hand, the structural (as opposed tofunctional) nature of HARDI tractography data enablesapplication of this imaging technique to both in vivo clinicalstudies and ex vivo neuropathologic studies.

Although HARDI and related techniques, such as dif-fusion spectrum imaging, are better able to resolve crossingfibers than diffusion tensor imaging (24, 26), there is currentlya limit at which the spatial and angular resolution, and hence

the sensitivity, of these techniques is exceeded. This meth-odological limitation is particularly relevant for in vivoHARDI studies because of the long duration of scan time thatis required to optimize the resolution of HARDI data. Wehave demonstrated that acquisition of HARDI data with suf-ficient resolution to identify ARAS connectivity is feasible ina healthy human subject, but human patients with neurologicdisease may not be able to tolerate long-duration HARDIscans (e.g. 30 minutes for Case 3). If HARDI data acquisitionparameters are not optimized because of limitations on theduration of scan time, it is possible that small fiber bundleswith complex branching patterns will evade detection byHARDI tractography, thereby producing false-negative con-nectivity results. For example, although the VTTR pathway tothe paraventricular region of the thalamus was identified byHARDI tractography in all 3 cases in this study, the complextrajectory of this pathway was more clearly delineated in Case1 than in Cases 2 and 3, likely because of the higher spatialresolution of the HARDI data (i.e. smaller voxel size) in theformer. Efforts are ongoing to shorten the duration of HARDIdata acquisition, improve spatial and angular resolution, andthereby facilitate clinical implementation (57). Yet, even withexpected improvements in HARDI data acquisition andpostprocessing, HARDI structural connectivity data should beinterpreted with caution and should not be considered as proofof functional synaptic connectivity.

Another methodological consideration is that our post-processing efforts to increase the specificity of ARAS tractidentification likely minimized, but did not completely elim-inate, the possibility that HARDI tractography could generatespurious tracts, that is, it may display a connection between2 fiber bundles that travel through or end within the samevoxel but that do not in fact connect (58). False-positiveconnectivity errors that may occur in HARDI tractographyexperiments may be analogous to leakage of an injected labelacross neuronal membranes into adjacent, nonrelevant cellsin histologic tract tracing studies (21). This methodologicaluncertainty is a particular concern when studying the brainstemcomponents of the ARAS, which arguably contain the greatestconcentration of complex crossing fibers of any brain region.

Despite these limitations, HARDI tractography has theability to generate connectivity maps of complex neural net-works in the postmortem human brain more effectively thancurrently available histologic labeling techniques, which areunable to identify long white matter tracts in adults (21). Inaddition, a significant advantage of HARDI tractography isthat it is a noninvasive technique that can be used to inves-tigate white matter connectivity in vivo and ex vivo. Indeed,in vivo tractography data from HARDI and diffusion spec-trum imaging studies have been correlated with known neu-roanatomic pathways in humans (24, 59), and ex vivotractography data from these techniques have been correlatedwith known neuroanatomic pathways in monkeys (24, 60), aswell as with histologic data in cats (61) and monkeys (60).Furthermore, the feasibility of applying HARDI tractographyto postmortem neuropathologic analyses is supported by evi-dence that the fixation of brain specimens does not alterHARDI data, even when the imaging is performed more than3 years after death. In a recent diffusion imaging study of






ownloaded from


11 human brain specimens that were fixed at an average of46.2 hours (range, 21Y69 hours) and imaged at an averageof 25.2 months (range, 2Y40 months) postmortem, the trac-tography results were consistent with those obtained fromin vivo human connectivity studies (43). Although quantitativewater diffusion measurements within white matter tracts maybe partially dependent on the time to fixation and the time toimage acquisition, there was no evidence that these alterationsaffected the reliability of the postmortem tractography results(43). Importantly, in this study, Case 1 was imaged 103 daysafter death and Case 2 was imaged 8 months 16 days afterdeath. Thus, imaging of both postmortem brain specimenswas performed well within the time window established byMiller et al (43).

Connectivity of the Cuneiform/SubcuneiformNucleus and Pontis Oralis

The major finding of our study is that HARDI trac-tography revealed a bilateral fiber bundle connecting thecuneiform/subcuneiform nucleus in the rostral midbrain to thethalamus, the hypothalamus, and the basal forebrain, provid-ing a structural basis in the human brain for the physiologicascending arousal pathway discovered by Moruzzi andMagoun (8) in cats. The originating source of this fiber bundlealso included the pontis oralis of the rostral pons, consistentwith its role in the ARAS based on clinicopathologic studies(6, 7). In the rostral midbrain, the bilateral fiber bundle orig-inating from the mesencephalic and pontine reticular formation,that is, the reticular core of the classic ARAS, bifurcates intothe VTT and DTT, names with historical precedent in exper-imental animals (46Y48). In animal studies, the ventral bundleis mainly a hypothalamic pathway, with divergent projectionsalso to the thalamic reticular nucleus, zona incerta, and Forelfields (47, 48). The human VTT, however, bifurcates yetagain into 2 additional fiber bundles: one labeled the VTTC,which is a distinct caudal pathway connecting with the hypo-thalamus, zona incerta, Forel fields, basal forebrain, and globuspallidus; and the other labeled the VTTR, which is a rostralpathway unexpectedly connecting with the paraventricular(midline) region of the thalamus.

The DTT has been shown in animals to project princi-pally to the thalamic intralaminar and paraventricular nucleibut not the reticular nucleus of the thalamus (47). In humans,however, we found that the DTT bifurcated into 2 additionalfiber bundles that each projected to a different distributionof thalamic nuclei. One subdivision, which we labeled theDTT lateral (DTTL), projected laterally to the thalamic retic-ular nucleus and, to a lesser extent, the basal forebrain, andthe pulvinar and lateral geniculate nucleus in the thalamus.The second subdivision of the DTT, which we labeled DTTmedial (DTTM), projected medially to thalamic intralaminarnuclei (central lateral nucleus and centromedian/paraf-ascicular nuclear complex). In addition, several DTTM fibersprojected beyond the intralaminar nuclei along a rostral andlateral course to the reticular nucleus.

Thus, the newly defined human VTTR, VTTC, DTTL,and DTTM pathways are distinct in their trajectory, con-nectivity, and source nuclei origins from the VTT and DTTpathways described in animals with tract tracing methods

(46Y48). We implicate the DTT and VTT pathways in themediation of arousal because they connect with thalamic,hypothalamic, and basal forebrain end-points well establishedto be crucial to this function. Our structural connectivityfindings therefore have significant implications for futurefunctional studies of human arousal, sleep-wake cycling, andconscious awareness. Similar to animal studies, this humanARAS analysis suggests a chain of connectivity wherebybrainstem source nuclei in the upper pons and midbrain projectto the intralaminar nuclei in the thalamus, which, in turn,project diffusely to the cerebral cortex to activate it (52),thereby enabling cognitive processing. In the human brain, wepostulate that the DTTM pathway to the intralaminar nuclei isthe main conduit for direct ARAS activation of thalamocort-ical networks (62). The human DTTL pathway, on the otherhand, connects to the thalamic reticular nucleus, which isknown to inhibit excitatory input to the cortex via connec-tions with the intralaminar nuclei and which produces thesynchronized 7- to 14-Hz oscillations on cortical EEG duringthe initial stage of sleep (17). Our tractography findings thussuggest a structural basis for the human ARAS to modulatearousal by both direct activation of thalamocortical networks(via DTTM) and by regulation of intrathalamic inhibitorynetworks (via DTTL).

Connectivity of the KnownNeurotransmitter-Specific Componentsof the ARAS

In this study, the finding that the known neurotransmitter-specific components of the ARAS connect in different pat-terns with the thalamus, hypothalamus, and basal forebrainunderscores the concept that these components underlie differ-ent behavioral and/or EEG aspects of the human arousal system.Indeed, each neurotransmitter-specific pathway is believed tobecome active in anticipation of arousal from sleep, yet withdistinct electrophysiological activation patterns (17). Theextensive connectivity between known brainstem mono-aminergic-, cholinergic-, and glutamatergic-related nuclei andthe anterior hypothalamus, which modulates circadian sleep-wake cycles (16), is consistent with animal data (63) andprovides support for current hypotheses that cyclical behav-ioral states are mediated by reciprocal interactions betweeninterconnected neurotransmitter networks (64). Furthermore,we provide a potential neuroanatomic basis in the humanbrain for visual imagery in the dream state, as the pathwayconnecting the pedunculopontine nucleus to the lateral geni-culate nucleus (essential to visual processing) is consistentwith evidence for cholinergic activation of REM sleep (65).

A potential limitation of HARDI tractography in map-ping the connectivity of the human ARAS is related to theobservation that our ex vivo connectivity analyses in Case 1did not identify a wide set of projections to the thalamus fromserotonergic- and noradrenergic-related source nuclei. Theseprojections have been suggested by in vivo human neuro-imaging studies with serotonergic and noradrenergic ligands(66, 67) and by postmortem neurochemical studies of thethalamus in animals and humans (68Y71). We did, however,identify extensive locus coeruleus and dorsal raphe con-nectivity with the thalamus (particularly with the reticular






ownloaded from


and centromedian/parafascicular nuclei) in Case 2 (ex vivo)and Case 3 (in vivo), suggesting that HARDI tractographycan indeed identify these important noradrenergic and sero-tonergic pathways in postmortem brain specimens and livingsubjects.

There are several potential explanations for these locuscoeruleus and dorsal raphe connectivity findings, which wereunexpected because the spatial resolution in Case 1 washigher than that in Cases 2 and 3 and because the angularresolution was similar in Cases 1 and 2. First, HARDI trac-tography may currently have a low sensitivity for identifyingtracts with multidirectional branching patterns (72), smallaxonal diameters (11), and lack of myelination (11)Vallknown characteristics of the rostral serotonergic and nora-drenergic axonal projections. Second, the formaldehyde fix-ative may interfere with MRI water diffusion measurementsalong poorly myelinated fiber tracts, although it is unclearwhy this potential fixative effect would impact the locuscoeruleus and dorsal raphe tractography results in Case 1more than in Case 2 because there was a shorter postmorteminterval before imaging in Case 1. Third, although the b val-ues used for ex vivo analysis in Case 1 and in vivo analysis inCase 3 were similar (4,057 and 4,000 s/mm2, respectively), itis possible that, at a given b value, the effective diffusioncontrast may be greater for in vivo experimental conditionsthan for ex vivo experimental conditions because global brainwater diffusivity is invariably lower in fixed postmortem tis-sue. In other words, it may be necessary to use even higher bvalues (i.e. 94,057 s/mm2) in studies of fixed brain tissue toreach the level of diffusion contrast that can reliably identifythe poorly myelinated locus coeruleus and dorsal raphe fibertracts. Additional studies are therefore needed to determinethe optimal HARDI data acquisition parameters (voxel size,number of diffusion gradients, and b value) for the identi-fication of each ARAS pathway ex vivo and in vivo. Fur-thermore, validation of neuroanatomic connectivity modelsgenerated by HARDI tractography in ex vivo and in vivostudies ultimately depends on correlation with anatomic tracttracing studiesVeither those already available in the literature(47, 60, 73) or, alternatively, in prospective experimentalstudies designed to confirm unexpected connections. In thisregard, the novel ARAS connections that we report herewarrant follow-up in neuroanatomic studies in human tissuesand/or animal models.

Intrathalamic ConnectivityThe intrathalamic connectivity findings in this study

provide a neuroanatomic basis for thalamic modulation, orgating, of the inputs of the ascending arousal system to thecerebral cortex (17, 52). Although we cannot exclude thepossibility that some fiber tracts passing between the intra-laminar nuclei and the reticular nuclei are thalamocorticalprojections passing through but not to the reticular nuclei,our finding that the reticular nucleus connects with both thecentral lateral nucleus and the centromedian/parafascicularnuclear complex is consistent with electrophysiological stud-ies in animals demonstrating specialization of thalamic intra-laminar nuclei and the existence of disynaptic intrathalamicpathways mediated by the reticular nucleus (74). Our obser-

vation that both the reticular nucleus of the thalamus and thebrainstem nuclei of the ARAS connect with Forel fields andzona incerta suggests an expanded role for these subthalamicregions as nodes in the neuroanatomic network mediatinghuman arousal, consistent with animal studies (75). Futurestudies in humans examining ARAS modulation of theactivity of the thalamic reticular nucleus may therefore con-sider the 3 distinct neuroanatomic pathways defined here: 1)the DTTL, 2) the intrathalamic extension of the DTTM, and 3)a potential monosynaptic pathway from the brainstem to Forelfields and zona incerta to the reticular nucleus.

In conclusion, the neuroanatomic connectivity results inthis study suggest that HARDI tractography has the potentialto facilitate the study of the human ARAS and its role inconsciousness, sleep-wake cycling, and coma (6, 7, 76, 77).The multiplicity and redundancy of the ascending arousalsystem suggest an adaptive mechanism for the recovery ofconsciousness when 1 component, but not the full system, isclinically disrupted (6). It remains to be determined which orhow many of the multiple components of the ARAS are suf-ficient for arousal, and hence consciousness, in humans. Theexamination of HARDI data from multiple patients with dif-ferent types of ARAS lesions and with different levels of con-sciousness may ultimately provide us with the answers to suchprovocative questions as which and how many connections areessential for human cognition. The precise identification offocal disruptions in arousal networks in the brainstem, thehypothalamus, the thalamus, and the basal forebrain of patientswith disorders of consciousness may also enable clinicians totarget individually tailored pharmacologic (78, 79) and elec-trophysiologic (e.g. deep brain stimulation [80]), therapies tospecific neuroanatomic sites of injury. In addition to its manyclinical applications, HARDI tractography can be used inneuropathologic studies to investigate the structural basis fordevelopmental differences in consciousness among the new-born, infant, child, adolescent, and adult, as well as the neu-roanatomic basis of coma, vegetative, and minimally consciousstates in pediatric and adult patients. We envision that post-mortem tractography will be a future adjunct to neuropath-ologic studies of patients with disorders of consciousness.Specific sites of disruption of ARAS pathways that are notappreciated on gross macroscopic analysis may be identifiedwith postmortem HARDI tractography, and these imaging datamay then be used to guide microscopic sectioning, therebyensuring precise clinicopathologic correlations. The exquisiteneuroanatomic delineation of the ARAS by HARDI tractog-raphy thus suggests that this advanced imaging technique hasthe potential to significantly advance the study of the physiol-ogy and pathology of human consciousness.

ACKNOWLEDGMENTSThe authors are grateful to Dr Martin A. Samuels for

his ongoing support of this work. The authors also thank MsMarian Slaney, Ms Michelle Siciliano, Ms Elisha Johnson,and Ms Allison Stevens for expert technical processing ofthe postmortem brain specimens; Dr Andre van der Kouwefor assistance with acquisition of in vivo HARDI data; andDr Jennifer A. McNab for helpful consultation regarding






ownloaded from


signal properties of the diffusion-weighted steady-state free-precession imaging sequence.

REFERENCES1. Kinney HC, Samuels MA. Neuropathology of the persistent vegetative

state. A review. J Neuropathol Exp Neurol 1994;53:548Y582. Steriade M. Arousal: Revisiting the reticular activating system. Science

1996;272:225Y263. Jones BE. Activity, modulation and role of basal forebrain cholinergic

neurons innervating the cerebral cortex. Prog Brain Res 2004;145:157Y69

4. McCormick DA. Neurotransmitter actions in the thalamus and cerebralcortex and their role in neuromodulation of thalamocortical activity. ProgNeurobiol 1992;39:337Y88

5. Parvizi J, Damasio A. Consciousness and the brainstem. Cognition 2001;79:135Y60

6. Parvizi J, Damasio AR. Neuroanatomical correlates of brainstem coma.Brain 2003;126:1524Y36

7. Posner JB, Saper CB, Schiff ND, et al. Plum and Posner’s Diagnosis ofStupor and Coma. 4th ed. New York, NY: Oxford University Press, 2007

8. Moruzzi G, Magoun HW. Brain stem reticular formation and activationof the EEG. Electroencephalogr Clin Neurophysiol 1949;1:455Y73

9. Steriade M, Glenn LL. Neocortical and caudate projections of intra-laminar thalamic neurons and their synaptic excitation from midbrainreticular core. J Neurophysiol 1982;48:352Y71

10. Starzl TE, Taylor CW, Magoun HW. Ascending conduction in reticularactivating system, with special reference to the diencephalon. J Neuro-physiol 1951;14:461Y77

11. Azmitia EC, Gannon PJ. The primate serotonergic system: A review ofhuman and animal studies and a report on Macaca fascicularis. AdvNeurol 1986;43:407Y68

12. Aston-Jones G, Cohen JD. Adaptive gain and the role of the locuscoeruleusYnorepinephrine system in optimal performance. J CompNeurol 2005;493:99Y110

13. Bromberg-Martin ES, Matsumoto M, Hikosaka O. Dopamine in moti-vational control: Rewarding, aversive, and alerting. Neuron 2010;68:815Y34

14. Fuller P, Sherman D, Pedersen NP, et al. Reassessment of the structuralbasis of the ascending arousal system. J Comp Neurol 2011;519:933Y56

15. Benarroch EE. Thermoregulation: Recent concepts and remaining ques-tions. Neurology 2007;69:1293Y97

16. Aston-Jones G, Chen S, Zhu Y, et al. A neural circuit for circadian reg-ulation of arousal. Nat Neurosci 2001;4:732Y38

17. Steriade M, McCormick DA, Sejnowski TJ. Thalamocortical oscillationsin the sleeping and aroused brain. Science 1993;262:679Y85

18. Jennett B, Plum F. Persistent vegetative state after brain damage. Asyndrome in search of a name. Lancet 1972;1:734Y37

19. Giacino JT, Ashwal S, Childs N, et al. The minimally conscious state:Definition and diagnostic criteria. Neurology 2002;58:349Y53

20. Alam M, Schwabe K, Krauss JK. The pedunculopontine nucleus area:Critical evaluation of interspecies differences relevant for its use as atarget for deep brain stimulation. Brain 2010;134:11Y23

21. von Bartheld CS, Cunningham DE, Rubel EW. Neuronal tracing withDiI: Decalcification, cryosectioning, and photoconversion for light andelectron microscopic analysis. J Histochem Cytochem 1990;38:725Y33

22. Tardif E, Clarke S. Intrinsic connectivity of human auditory areas: Atracing study with DiI. Eur J Neurosci 2001;13:1045Y50

23. Kinomura S, Larsson J, Gulyas B, et al. Activation by attention of thehuman reticular formation and thalamic intralaminar nuclei. Science1996;271:512Y15

24. Wedeen VJ, Wang RP, Schmahmann JD, et al. Diffusion spectrummagnetic resonance imaging (DSI) tractography of crossing fibers.Neuroimage 2008;41:1267Y77

25. Tournier JD, Calamante F, Gadian DG, et al. Direct estimation of thefiber orientation density function from diffusion-weighted MRI datausing spherical deconvolution. Neuroimage 2004;23:1176Y85

26. Tuch DS, Reese TG, Wiegell MR, et al. Diffusion MRI of complexneural architecture. Neuron 2003;40:885Y95

27. Basser PJ, Pierpaoli C. Microstructural and physiological features oftissues elucidated by quantitative-diffusion-tensor MRI. J Magn Reson B1996;111:209Y19

28. Pierpaoli C, Basser PJ. Toward a quantitative assessment of diffusionanisotropy. Magn Reson Med 1996;36:893Y906

29. Shapiro EM, Skrtic S, Sharer K, et al. MRI detection of single particlesfor cellular imaging. Proc Natl Acad Sci U S A 2004;101:10901Y6

30. McNab JA, Miller KL. Sensitivity of diffusion weighted steady state freeprecession to anisotropic diffusion. Magn Reson Med 2008;60:405Y13

31. McNab JA, Jbabdi S, Deoni SC, et al. High resolution diffusion-weightedimaging in fixed human brain using diffusion-weighted steady state freeprecession. Neuroimage 2009;46:775Y85

32. Buxton RB. The diffusion sensitivity of fast steady-state free precessionimaging. Magn Reson Med 1993;29:235Y43

33. Reese TG, Heid O, Weisskoff RM, et al. Reduction of eddy-current-in-duced distortion in diffusion MRI using a twice-refocused spin echo.Magn Reson Med 2003;49:177Y82

34. Wang R, Wedeen VJ. Diffusion Toolkit, Version 6.0 and TrackVis,Version 5.1. Charlestown, MA: Athinoula A. Martinos Center forBiomedical Imaging, Massachusetts General Hospital, 2011. Availableat: http://www.trackvis.org

35. Paxinos G, Huang X. Atlas of the Human Brainstem. San Diego, CA:Academic Press, 1995

36. Olszewski J, Baxter D. Cytoarchitecture of the Human Brainstem. 2nded. Basel, Switzerland: Karger, 1982

37. Morel A, Magnin M, Jeanmonod D. Multiarchitectonic and stereotacticatlas of the human thalamus. J Comp Neurol 1997;387:588Y630

38. Catani M, Howard RJ, Pajevic S, et al. Virtual in vivo interactive dis-section of white matter fasciculi in the human brain. Neuroimage 2002;17:77Y94

39. Hess CP, Mukherjee P, Han ET, et al. Q-ball reconstruction of multi-modal fiber orientations using the spherical harmonic basis. Magn ResonMed 2006;56:104Y17

40. Mori S, Crain BJ, Chacko VP, et al. Three-dimensional tracking ofaxonal projections in the brain by magnetic resonance imaging. AnnNeurol 1999;45:265Y69

41. Newcombe VF, Williams GB, Scoffings D, et al. Aetiological differencesin neuroanatomy of the vegetative state: Insights from diffusion tensorimaging and functional implications. J Neurol Neurosurg Psychiatry2010;81:552Y61

42. Wiegell MR, Tuch DS, Larsson HB, et al. Automatic segmentation ofthalamic nuclei from diffusion tensor magnetic resonance imaging.Neuroimage 2003;19:391Y401

43. Miller KL, Stagg CJ, Douaud G, et al. Diffusion imaging of whole,post-mortem human brains on a clinical MRI scanner. Neuroimage 2011;57:167Y81

44. Salat DH, Tuch DS, Greve DN, et al. Age-related alterations in whitematter microstructure measured by diffusion tensor imaging. NeurobiolAging 2005;26:1215Y27

45. Miller KL, McNab JA, Jbabdi S, et al. Diffusion tractography of postmortem human brains: Optimization and comparison of spin echo andsteady-state free precession techniques. Neuroimage 2011;59:2284Y97

46. Shute CC, Lewis PR. Cholinesterase-containing systems of the brain ofthe rat. Nature 1963;199:1160Y64

47. Vertes RP, Martin GF. Autoradiographic analysis of ascending projec-tions from the pontine and mesencephalic reticular formation and themedian raphe nucleus in the rat. J Comp Neurol 1988;275:511Y41

48. Nauta WJH, Kuypers HGJM. Some ascending pathways in the brain stemreticular formation. In: Jasper HH, Proctor LD, Knighton RS, NoshayWC, Costello RT, eds. Reticular Formation of the Brain. Boston, MA:Little, Brown, and Company, 1958:3Y30

49. Yoshiura T, Wu O, Zaheer A, et al. Highly diffusion-sensitized MRIof brain: Dissociation of gray and white matter. Magn Reson Med 2001;45:734Y40

50. Pierpaoli C, Jezzard P, Basser PJ, et al. Diffusion tensor MR imaging ofthe human brain. Radiology 1996;201:637Y48

51. Mulkern RV, Haker SJ, Maier SE. On high b diffusion imaging in thehuman brain: Ruminations and experimental insights. Magn ResonImaging 2009;27:1151Y62

52. Benarroch EE. The midline and intralaminar thalamic nuclei: Anatomicand functional specificity and implications in neurologic disease. Neu-rology 2008;71:944Y49

53. Basser PJ, Pierpaoli C. A simplified method to measure the diffusiontensor from seven MR images. Magn Reson Med 1998;39:928Y34






ownloaded from


54. Zhan L, Leow AD, Jahanshad N, et al. How does angular resolutionaffect diffusion imaging measures? Neuroimage 2010;49:1357Y71

55. Melhem ER, Itoh R, Jones L, et al. Diffusion tensor MR imaging of thebrain: Effect of diffusion weighting on trace and anisotropy measure-ments. AJNR Am J Neuroradiol 2000;21:1813Y20

56. Axer M, Amunts K, Grassel D, et al. A novel approach to the humanconnectome: Ultra-high resolution mapping of fiber tracts in the brain.Neuroimage 2011;54:1091Y101

57. Reese TG, Benner T, Wang R, et al. Halving imaging time of whole braindiffusion spectrum imaging and diffusion tractography using simulta-neous image refocusing in EPI. J Magn Reson Imaging 2009;29:517Y22

58. Basser PJ, Pajevic S, Pierpaoli C, et al. In vivo fiber tractography usingDT-MRI data. Magn Reson Med 2000;44:625Y32

59. Chao YP, Cho KH, Yeh CH, et al. Probabilistic topography of human corpuscallosum using cytoarchitectural parcellation and high angular resolutiondiffusion imaging tractography. Hum Brain Mapp 2009;30:3172Y87

60. Schmahmann JD, Pandya DN, Wang R, et al. Association fibre pathwaysof the brain: Parallel observations from diffusion spectrum imaging andautoradiography. Brain 2007;130:630Y53

61. Takahashi E, Dai G, Rosen GD, et al. Developing neocortex organizationand connectivity in cats revealed by direct correlation of diffusion trac-tography and histology. Cereb Cortex 2011;21:200Y11

62. Morison RS, Dempsey EW. A study of thalamo-cortical relations. Am JPhysiol 1942;135:281Y92

63. Krout KE, Kawano J, Mettenleiter TC, et al. CNS inputs to the supra-chiasmatic nucleus of the rat. Neuroscience 2002;110:73Y92

64. Hobson JA, McCarley RW, Wyzinski PW. Sleep cycle oscillation: Recip-rocal discharge by two brainstem neuronal groups. Science 1975;189:55Y58

65. Hobson JA. REM sleep and dreaming: Towards a theory of protocon-sciousness. Nat Rev Neurosci 2009;10:803Y13

66. Takano H, Ito H, Takahashi H, et al. Serotonergic neurotransmission inthe living human brain: A positron emission tomography study using[11C]dasb and [11C]WAY100635 in young healthy men. Synapse 2011;65:624Y33

67. Ding YS, Singhal T, Planeta-Wilson B, et al. PET imaging of the effectsof age and cocaine on the norepinephrine transporter in the human

brain using (S,S)-[11C]O-methylreboxetine and HRRT. Synapse 2010;64:30Y38

68. Varnas K, Halldin C, Hall H. Autoradiographic distribution of serotonintransporters and receptor subtypes in human brain. Hum Brain Mapp2004;22:246Y60

69. Oke A, Keller R, Mefford I, et al. Lateralization of norepinephrine inhuman thalamus. Science 1978;200:1411Y13

70. Rodriguez JJ, Noristani HN, Hoover WB, et al. Serotonergic projectionsand serotonin receptor expression in the reticular nucleus of the thalamusin the rat. Synapse 2011;65:919Y28

71. Vogt BA, Hof PR, Friedman DP, et al. Norepinephrinergic afferents andcytology of the macaque monkey midline, mediodorsal, and intralaminarthalamic nuclei. Brain Struct Funct 2008;212:465Y79

72. Grzanna R, Molliver ME. Cytoarchitecture and dendritic morphology ofcentral noradrenergic neurons. In: Hobson JA, Brazier MAB, eds. TheReticular Formation Revisited: Specifying Function for a NonspecificSystem. New York, NY: Raven Press, 1980:83Y97

73. Schmahmann JD, Pandya DN. Fiber Pathways of the Brain. New York,NY: Oxford University Press, 2006

74. Crabtree JW, Isaac JT. New intrathalamic pathways allowing modality-related and cross-modality switching in the dorsal thalamus. J Neurosci2002;22:8754Y61

75. Trageser JC, Burke KA, Masri R, et al. State-dependent gating of sensoryinputs by zona incerta. J Neurophysiol 2006;96:1456Y63

76. Brown EN, Lydic R, Schiff ND. General anesthesia, sleep, and coma.N Engl J Med 2010;363:2638Y50

77. Owen AM, Schiff ND, Laureys S. A new era of coma and consciousnessscience. Prog Brain Res 2009;177:399Y411

78. Fridman EA, Krimchansky BZ, Bonetto M, et al. Continuous subcuta-neous apomorphine for severe disorders of consciousness after traumaticbrain injury. Brain Inj 2010;24:636Y41

79. Giacino JT, Whyte J, Bagiella E, et al. Placebo-controlled trial of amanta-dine for severe traumatic brain injury. N Eng J Med 2012;366:819Y26

80. Schiff ND, Giacino JT, Kalmar K, et al. Behavioural improvements withthalamic stimulation after severe traumatic brain injury. Nature 2007;448:600Y3






ownloaded from


Neuroanatomic Connectivity of the Human Ascending Arousal ...ORIGINAL ARTICLE Neuroanatomic...

Documents

Transcript of Neuroanatomic Connectivity of the Human Ascending Arousal ...ORIGINAL ARTICLE Neuroanatomic...