A network approach to assessing cognition in disorders of consciousness

Abstract

Objective:

Conventional assessments of consciousness rely on motor responses to indicate awareness. However, overt behaviors may be absent or ambiguous in patients with disorders of consciousness (DOC) resulting in underrating capacity for cognition. fMRI during a silent picture-naming task was evaluated as an indicator of command following when conventional methods are not sufficient.

Methods:

A total of 10 patients with and without conventional evidence of awareness, who met diagnostic criteria for the minimally conscious state (MCS) (n = 5), vegetative state (VS) (n = 3), emerged from MCS (EMCS) (n = 1), and locked-in syndrome (LIS) (n = 1), participated in this observational fMRI study.

Results:

The LIS and EMCS patients engaged a complete network of essential language-related regions during the object-naming task. The MCS and 2 of the VS patients demonstrated both complete and partial preservation of the object-naming system. Patients who engaged a complete network scored highest on the Coma Recovery Scale-Revised.

Conclusions:

This study supports the view that fMRI during object naming can elicit brain activations in patients with DOC similar to those observed in healthy subjects during command following, and patients can be stratified by completeness of the engaged neural system. These results suggest that activity of the language network may serve as an indicator of high-level cognition and possibly volitional processes that cannot be discerned through conventional behavioral assessment alone.

GLOSSARY

BA

= Brodmann area;

BOLD

= blood oxygenation level–dependent;

CRS-R

= Coma Recovery Scale-Revised;

DOC

= disorders of consciousness;

EMCS

= emerged from minimally conscious state;

GFi(d)

= dorsal inferior frontal gyrus;

GFi(v)

= ventral inferior frontal gyrus;

hrf

= hemodynamic response function;

LIS

= locked-in syndrome;

MCS

= minimally conscious state;

preSMA

= pre-supplementary motor area;

STG

= superior temporal gyrus;

VS

= vegetative state.

Awareness of self or environment is a component of conscious processing and is conventionally assessed by overt motor responses. Using this method, patients who are in minimally conscious state (MCS) demonstrate fluctuations in awareness and clear, but inconsistent, minimal behavioral signs of conscious awareness,¹ and patients in vegetative state (VS) are awake but assumed not aware, because they do not show behavioral responses to environmental stimuli in a contingent manner.² On the other hand, patients with LIS (locked-in syndrome) similarly show very limited signs of awareness due to profound sensory and motor deficits but have preserved self-awareness and normal or near-normal cognitive capacities.³ Thus, impairments of the sensory and motor systems in these patient populations may render conventional bedside diagnosis inadequate and are thought to have contributed to high rates of misdiagnosis of LIS, MCS, and VS⁴ as well as underestimation of cognitive capacity. Recent neuroimaging studies using passive tasks,^5–7 imagery tasks,^8–11 and covert tasks^12,13 have suggested that a larger number of patients with disorders of consciousness (DOC) than previously suspected may retain some level of awareness undetected by conventional behavioral response methods.

The goal of this study was to evaluate a silent picture-naming task as a command-following paradigm to assess putative cognitive capacity in patients who fail to demonstrate discernible evidence of awareness on bedside examination. We considered the specificity of brain activity to this task to be a representative index of covert responses by virtue of the distinction between patterns of activity known to be elicited by this task and those elicited by passive viewing.^14–17

METHODS

Subjects.

All patients entering the brain trauma unit at JFK Johnson Rehabilitation Institute were screened for study eligibility over the course of 3 years. All potentially eligible patients were also recruited at the Weill Cornell Medical Center–New York Presbyterian Hospital. Initial recruitment excluded patients with MRI incompatible devices, contraindications for ambulance transportation, evidence of lesions involving two-thirds or more of Wernicke, Broca, or both areas, and uncontrolled seizure disorders. Seventeen consecutive patients were recruited for this study. Here we report 10 right-handed patients, 5 male, age 18–58 years (mean = 34.4 ± 15.9) (table 1). Seven patients were not included due to 1) severe motion artifacts (1 patient in MCS and 1 patient in EMCS), 2) inability to complete the 2 picture-naming runs due to fatigue, lack of fixation on stimulus, or restlessness (2 patients in MCS and 1 patient in EMCS), 3) severe imaging artifacts due to abnormal brain structure encompassing half or more of an entire hemisphere (1 patient in VS), or 4) known visual impairments (1 patient in MCS). One patient was diagnosed as in LIS according to Haig et al.,³ 1 patient met the diagnostic criteria for EMCS,¹ 5 patients met the criteria for MCS,¹ and 3 patients met the criteria for VS set by the American Academy of Neurology.¹⁸ At the time of the scan, the patient with LIS had recovered lateral head movement, right-hand finger movements, vertical eye movements, and consistent, accurate communication.

Table 1 Summary of demographic and clinical information

Standard protocol approvals and patient consents.

Signed assent from the patient's legal representative was obtained prior to participation in the study in accordance with established institutional guidelines under protocol AAAA-3690. This study was approved by the Institutional Review Boards of Columbia University, the Solaris Health System, and Weill Cornell Medical Center–New York Presbyterian Hospital.

Imaging.

All scans were collected on a 1.5-T General Electric (LX) magnetic resonance scanner with a standard head coil. Functional scans consisted of 2 axial EPI time series (128 × 128 mm matrix, 21 contiguous slices, field of view = 190 mm, 1.5 × 1.5 × 5 mm voxels, repetition time = 3,000 msec, echo time = 43 msec, flip angle = 60^o) obtained parallel to the AC/PC line. A total of 48 images was acquired in each run. Each active condition lasted 40 seconds and was preceded and followed by a resting period of 40 seconds. Conventional high-resolution (T1-weighted) images were also acquired along sagittal planes and at the same axial plane locations as the T2*-weighted images.

fMRI paradigm.

The fMRI activation paradigm consisted of 2 separate block runs. The visual stimuli consisted of drawings of objects (Boston Naming Set 2), and stimuli were presented on MRI-compatible goggles (n = 3) or back-projected on a screen (n = 7) depending on patient tolerance. Eye tracking was not possible, as patients are required to perform eye movements to specific locations on command for calibration. However, when the goggles were employed, eye fixation was confirmed by inspection of the pupil as shown on the monitor. When back-projection was employed, fixation was monitored by an investigator in the scanner room. Patients were instructed to look at the screen, rest during the presence of a “plus sign,” and silently name the objects as they appeared. The following instructions were given to the patients: “Whenever you see a cross sign, just relax and rest. After a while, you will see pictures of objects on the screen. When you see these objects, name then to yourself.” The instructions for the task were given verbally before each run by an investigator through headphones designed to reduce background scanner noise (Resonance Technology, Inc.). In the case of LIS, all instructions were written to accommodate previously diagnosed auditory processing deficits. Patients performed additional auditory and visual passive paradigms as part of a larger study.

Statistical analysis.

Image processing and analysis was performed with SPM2 software (Wellcome Department of Imaging Neuroscience, University College, London, UK). Preprocessing steps included motion correction using a sync interpolation algorithm, and smoothing with a 6-mm full-width-half-maximum Gaussian kernel. The blood oxygenation level–dependent (BOLD) response was represented by a boxcar function for the active epoch that was convolved with the canonical hemodynamic response function (hrf). A 160-second high-pass filter was used to remove low-frequency confounds and an hrf-shaped low-pass filter was used to remove unknown temporal autocorrelations. Motion parameters were included as regressors of no interest. In addition, large motion events were modeled with scan nulling regressors as previously described for epileptic patients.¹⁹ Statistical analysis was based on the general linear model. The design matrices modeled the 2 runs separately. Contrasts were aimed at the identification of activations related to language activity corresponding to signal intensities greater than the baseline. Given that signal amplitude may be reduced in this patient population as in patients with vascular disorders,²⁰ and all results are reported for each individual patient, we selected a statistical threshold of t ≥2.35 (p = 0.01 uncorrected), intended to avoid possible false-negative findings. A minimum cluster size of 10 voxels was used as an extent threshold (k). A similar approach is used with some clinical populations where signal variability is large across subjects.²¹ For each subject, labeling was performed using the whole-brain analysis. Assignments of anatomic labels for active regions were based on correspondence between the patient's brain anatomy and the human brain atlas as previously described.⁵

Behavioral assessment.

The Coma Recovery Scale-Revised (CRS-R)²² was used for diagnoses. CRS-R consists of 23 items organized in 6 subscales that address arousal, auditory, visual, motor, oromotor/verbal, and communication systems. Each subscale is organized hierarchally, with lower items representing reflexive activity and higher items representing cognitive-based behaviors.

RESULTS

The picture-naming task is designed to target a network of regions specialized for language functions including Broca area and Wernicke area and medial frontal gyrus (generally on the left hemisphere). Our previous results using the same task in 3 sensory modalities in a group of healthy subjects^21,23 show activity in the bilateral primary visual and association cortices, middle/superior temporal gyrus, left ventral/dorsal inferior frontal gyrus, and medial frontal cortex. We use the reported supramodal pattern of functional activation as the normative areas for internal picture naming. The extent to which the patients in this study show activation in the normative brain regions across all diagnostic categories is presented in table 2 and figure 1.

Table 2 Functional activation of normative areas in each patient for the internal picture-naming task and degree of preservation of the naming network

Figure 1 fMRI results for the covert picture naming for individual patients

Colored voxels indicate blood oxygenation level–dependent (BOLD) responses elicited by the task. Activations are overlaid on each patient's brain and shown in a representative slice for each anatomic area (top row). Color bars reflect the range of BOLD signal intensity changes with corresponding t values. EMCS = emerged from minimally conscious state; LIS = locked-in syndrome; MCS = minimally conscious state; VS = vegetative state.

Because the picture-naming task is a visual task, the presence of visual cortex activation is taken as a prerequisite for successful performance of the task, and all patients reported here activated the visual cortex in response to the visual stimulation. LIS, EMCS, and MCS4 activated the primary visual cortex (p = 0.001) bilaterally, and the remaining patients engaged primary visual cortex at p = 0.01 in at least one hemisphere, suggesting that all of the subjects performed the visual encoding of the pictures required for the picture-naming task (table e-1 on the Neurology® Web site at www.neurology.org). In addition to engagement of early visual cortex, the middle occipital and fusiform gyri associated with object recognition²⁰ were active in all the patients except VS3.

Activation of left temporal regions expected during naming, which mediate semantic knowledge of the object^24–26 associated with object recognition, was observed in LIS, EMCS, MCS1–MCS5, VS2, and VS3 (STG, table 2 and figure 1). Right temporal activation was also present in EMCS, MCS1, and MCS3, and VS1 showed only right temporal activation.

With the exception of VS1, the left ventral inferior frontal gyrus, thought to be associated with phonologic processing,^26–28 more specifically with retrieval of the object's name, was activated in all patients (GFi(v), table 2 and figure 1). In addition, LIS, EMCS, MCS1, MCS2, MCS3, VS1, and VS3 activated the right ventral inferior frontal gyrus. Left dorsal inferior frontal gyrus [GFi(d), table 2 and figure 1] is reported to be related to selection of lexical representations²⁹ and preparation for overt articulation of language³⁰ and was engaged by LIS, EMCS, MCS1, MCS2, MCS3, MCS4, VS2, and VS3. MCS1 also activated the right hemisphere GFi(d), while VS1 only activated the right hemisphere GFi(d).

Engagement of the pre-supplementary motor area (preSMA) in the medial frontal gyrus, reportedly related to cognitive control³¹ and volitional aspects of cognition,³² was noted in patients with the highest CRS-R scores: LIS, EMCS, MCS1, and MCS2. An exception to this pattern was VS2, who showed activation in left preSMA. We also observed activation of superior frontal gyrus in LIS, MCS1, and VS1 (table e-1).

Our data demonstrate variations in underlying brain activity of the MCS patients for the command-following task (figure 2). The comparison of the naming network observed in healthy volunteers with our patients with DOC suggests a functional differentiation of patients into 3 groups: 1) complete network: patients with activation of all brain areas observed in healthy volunteers; 2) partial network: patients with activation of some brain areas observed in healthy volunteers; and 3) absent network: patients with no activation of any brain areas observed in healthy volunteers. This classification is shown in the last column of table 2. Our findings show that these neural categories can be generally related to behavioral measures of status. As expected, patients with high CRS-R scores, including LIS, EMCS, MCS1, and MCS2, show a pattern of activity similar to those of healthy subjects elicited by picture-naming task (complete network group). Conversely, patients with low CRS-R scores, i.e., VS1, show a very unresponsive brain as expected for this population (absent network group). Remarkably, patients in the partial network group are characterized by lack of activation in the medial frontal gyrus and lower CRS-R scores, i.e., MCS3 to MCS5.

Figure 2 Relationship between Revised Coma Recovery Scale (CRS-R) scores and preservation of the picture-naming network

Patients with disorders of consciousness are grouped according to preservation of the naming network previously observed in healthy volunteers: complete network: patients with activation of all brain areas observed in healthy volunteers (left superior temporal gyrus, left ventral inferior frontal gyrus, left dorsal inferior frontal gyrus, and pre-supplementary motor area); partial network: patients with activation of some brain areas observed in healthy volunteers; and 3) absent network: patients with no activation of any brain areas observed in healthy volunteers. The bars show the average CRS-R score for each group with the corresponding SD.

Nonetheless, we do not observe a strict correspondence between behavioral profile and overall brain activation in agreement with previous studies of language comprehension in patients with DOC.^6–7 Two patients in VS present evidence for some preserved picture-naming activity similar to controls (VS2 in the complete network group and VS3 in the partial network group).

DISCUSSION

In this study, we employed fMRI during covert picture naming (internal speech) as a command-following task in a series of patients with a range of DOC and behavioral profiles. We observed complete networks expected for the object-naming task in the LIS patient and the ECMS patient, in 2 of the 5 MCS patients, and 1 of the 3 VS patients. These findings demonstrate apparent preservation of some command-following capacity in patients with minimal or no signs of awareness as evidenced by the conventional behavioral examination. Nonetheless, even with these exceptions, we demonstrate a general correspondence between the integrity of the language-specific network and the bedside assessment (CRS-R). These data suggest that fMRI during object-naming tasks may provide additional insight regarding cognitive status for patients with compromised ability to respond by conventional methods.

Picture naming is considered a complex process that recruits both stimulus-driven (bottom-up)³³ and purpose-driven (top-down)³⁴ pathways including visual input, language, short-term memory, and motor and executive functions. Using a supramodal approach, we previously reported that this task recruited activity in the superior temporal gyrus Brodmann area (BA) 22 (including Wernicke area), inferior frontal gyrus BA 44/45 (including Broca area), and medial frontal gyrus (pre-SMA, BA 6/8).^21,23 Similar patterns of activity have also been reported by other groups and including parietal regions BA 39/40.^35–37 These regions are thought to form a complete long-range network during picture naming. The left BA 22 of the superior temporal gyrus, also referred to as Wernicke area, has been associated with semantic knowledge^24–26 and phonologic code retrieval³⁶; the left ventral region of the inferior frontal gyrus, BA 44/45, has been linked to phonologic processing,²⁷ selection of competing semantic representations,²⁸ and phonetic sequencing³⁷; and dorsal BA 44/46 has been associated with executive processes to select a lexical representation²⁹ and processes related to motor programming and planning of articulation.³⁰ Previous imaging reports have associated frontal midline regions with cognitive control³¹ and volitional actions.^9,12,32 Although emphasis has been commonly placed in the linguistic aspects of picture naming, here we used it as a probe of command following.

Although passive viewing of objects is known to recruit occipital and temporal lobe areas,³⁸ the active task of naming adds an additional load to the neural circuitry that engages other frontal regions and presumptive attention-related resources. In this study, we observed the expected automatic perceptual and object recognition areas during viewing plus the naming-related areas that involve the use of controlled attentional resources.³⁴ Our observations are consistent with previous EEG, magnetoencephalography, and fMRI studies that have reported brain signals limited to occipital and temporal areas associated with object perception and recognition during passive viewing, and have shown that activity extends to frontal regions during covert and overt picture naming distinguishing the active task from passive viewing.^14–16

The comparison of naming networks with patients and healthy volunteers allows a classification of patients into 3 categories: patients who retain all components of the reference network (complete network), patients who engage only some of those brain regions (partial network), and patients who show no activation of any of those components (absent network). For the MCS cohort, activation of neural networks associated with language contrasts with the lack of consistent verbal expression in these patients. The results for the VS cohort suggest that even behaviorally unresponsive patients who typically score low in the CRS-R can show signs of awareness using this technique. The finding of unresponsiveness at the behavioral level accompanied by extensive preservation of cognitive networks, as in VS2, challenges our convention to identify VS by behavioral assessment alone in some patients and documents a need for further evaluation of these patients.

Given that picture naming is a nonautomatic (effortful) process,³⁹ the observation of expected patterns of activity suggests preserved command-following capacities in these patients. The general correspondence to CRS-R measures suggest that neural correlates of object naming can be used as an index of cognitive capacity when motor responses are not sufficient. The engagement of the naming network may suggest at least partial preservation of purposeful behavior. However, we cannot rule out that neural reorganization in some cases may have occurred outside the naming network in patients who have been injured long before the study. The full extent to which the volitional capacity of patients with DOC can be equated to the experience of awareness in healthy volunteers remains a topic for further investigation.

The high proportion of patients who retain brain activations suggestive of command-following capacity (3 of the 8 patients who remain in a DOC state) in our study in comparison to previous studies using other imagery tasks may suggest that the recovery of language capacities antecedes other observed brain or behavioral responses. Alternatively, it could be that the use of visual activity as inclusion criterion biases our sample toward patients with this ability or that motor-compromised patients have more difficulty with gross motor imagery tasks. Importantly, assessing the integrity of the expressive language system may provide a proxy for the capacity to harness assistive communication systems, although our data do not directly address this possibility.

Command-following paradigms have a number of inherent limitations. All of these tasks require coordination of multiple cognitive systems (visual, auditory, language, working memory) and the absence of brain response could be due to damage or lack of recruitment of one or more of these systems. The functional capacity of these systems can be influenced by anxiety and fluctuations in arousal and awareness. In addition to these caveats, fMRI data acquisition may be compromised by head motion, and interpretation of imaging data can be challenged by the presence of structural abnormalities, a temporal mismatch between presentation of the target stimulus and the patient's response, or BOLD signal characteristics different from those of healthy controls.

Information provided by fMRI and an object-naming task as an indicator of command following may be indicated particularly in cases with movement impairments. These findings, particularly in the case of VS2, indicate a clear need to consider the capacity for covert cognitive processing in addition to the overt behaviors when conducting diagnostic assessment. Determining where patients lie on the spectrum of consciousness is key to optimal clinical management, and delivery of available therapies for maximization of quality of life for this population. fMRI and the network assessment approach presented here provides additional information regarding the potential cognitive capabilities of such patients, thereby offering a novel means of assessing cognitive status that cannot be documented by traditional methods.

ACKNOWLEDGMENT

The authors thank the patients and their families for participation in the study.

DISCLOSURE

Dr. Rodriguez Moreno has received fellowship support from the NIDRR (H133A070030). Dr. Schiff serves on a scientific advisory board for IntElect Medical, Inc.; is listed as coinventor on numerous patents re: Deep brain stimulation technology for which he receives royalties via Cornell University from IntElect Medical, Inc.; receives royalties from the publication of Plum and Posner's Diagnosis of Stupor and Coma, 4th ed. (Oxford University Press, 2007); and receives research support from IntElect Medical, Inc., the NIH (NINDS 1R01NS067249 [PI] and NICHD 1RO1HD51912 [PI]), and the James S. McDonnell Foundation. Dr. Giacino serves on the editorial board of the Journal of Head Trauma Rehabilitation; has served as a training consultant for IntElect Medical, Inc.; has received research support from IntElect Medical, Inc., the NIDRR (H133A031713 [PI] and H133A070030 [Local Project PI]); and has served as an expert medico-legal consultant. Dr. Kalmar and Dr. Hirsch report no disclosures.

Address correspondence and reprint requests to Dr. D. Rodriguez Moreno or Dr. Hirsch, Department of Radiology, Functional MRI Research Center, Neurological Institute B41, Box 108, 710 West 168th Street, New York, NY 10032 drmoreno2002@yahoo.com or joyhirsch@yahoo.com

Editorial, page 1860

Supplemental data at www.neurology.org

e-Pub ahead of print on October 27, 2010, at www.neurology.org.

Study funding: Supported by the NIDRR (H133A070030), the NIH (RO1NS056274), and James S. McDonnel Foundation.

Disclosure: Author disclosures are provided at the end of the article.

Received February 12, 2010. Accepted in final form July 8, 2010.