Cerebral response to subject’s own name showed high prognostic value in traumatic vegetative state

Overall, 74 DOC patients with severe brain injury were included in this study. After fMRI scanning, 8 were excluded due to head movement parameters (after scanning, patients’ head movement parameters were extracted and excluded from the threshold list as: translation >8 mm or roll angle >2°). Of the remaining 66 patients (52 male, 14 female, age range 2 to 73 years, mean 39 years; etiology: 43 traumatic, 12 anoxic brain injury, 10 cerebrovascular accident, 1 meningitis; time between ictus and fMRI scanning: 1 to 60 months, mean 8.5 months), 39 patients met the diagnostic criteria defining VS/UWS (23 traumatic and 16 non-traumatic), 25 patients met the diagnostic criteria defining MCS (19 traumatic and 6 non-traumatic), and 2 patients were diagnosed as EMCS (1 traumatic and 1 non-traumatic) according to the Coma Recovery Scale-Revised (CRS-R) [12] (see Table 1 for detailed demographic and clinical information). Data on 11 of the 66 patients (VS/UWS: 24, 25, 26, 27, 28, 29, and 30; MCS: 14, 15, 16, and 17) has previously been reported [6].

All patients were admitted to the rehabilitation units of Hangzhou Wujing hospital and Taizhou municipal hospital. Prior to admission, all patients were verified suitable for MRI scanning by a team of expert neurologists. Patients who had suffered brain injury less than one month before the time of evaluation were excluded from the study. An observation of patients’ baseline head movement was also performed at bedside by caregivers to check the patients were able to stay in the scanner for at least 10 minutes without excessive head movement. Informed written consent was obtained from the physician and family of each patient. Written informed consent for the publication of patient details was also obtained from the legal representative of all patients. The study was approved by the Ethics Committee of Hangzhou Normal University School of Basic Medicine.

Fifteen volunteer college undergraduate students (9 female, age range 18 to 27 years, mean age 24 years) participated in the study as healthy controls. All the volunteers had normal hearing and normal or corrected-to-normal vision. None reported any history of head injury or neurological or psychiatric disorders. Written informed consent was obtained from each participant prior to the experiment according to a protocol approved by the Ethics Committee of Hangzhou Normal University School of Basic Medicine.

All patients recruited to the study underwent behavioral assessment employing the CRS-R before fMRI data acquisition. In order to obtain a significant prognostic value, a follow-up program was conducted at 3, 6, and 12 months after fMRI data acquisition. If the patients were discharged from or transferred to other hospitals during this tracking program, a phone call follow-up was performed. The assessors for determining outcome were blinded for the fMRI results.

Before acquiring neuroimaging data, we digitally recorded and adapted the SON-FV using the voice of a first-degree relative and GoldWave software (GoldWave Inc.). fMRI scanning was performed using block design with six active blocks and seven baseline blocks for each patient. Each active block lasted 12 seconds and included seven SON-FVs (each name lasted 1 sec). Each baseline block consisted of 18 seconds of attenuated machine noise; 13 patients were scanned twice in the same day (MRI scanner and conditions were the same in both cases) to obtain the overlap rate of two separate scans with the same fMRI paradigm. The auditory stimuli were presented through MRI-compatible noise-attenuated headphones (Resonance Technology, Inc., Los Angeles, CA) binaurally. A special designed head-fixation devise (a pumping pillow) and polystyrene foam was used for every patient to reduce spontaneous head movement. Data were acquired using a 1.5 T General Electrics Sigma Horizon MRI system and a 1.5 T Siemens Magnetom Essenza MRI system (15 controls, VS/UWS1-30 and MCS1-17using GE MRI; VS/UWS31-39, MCS17-25, and EMCS1-2 using Siemens MRI). When scanning, first, 22 axial anatomic images were collected using a T1-weighted spin echo sequence (repetition time = 500 msec, echo time = 9 msec, field of view = 240 × 240 mm, slice thickness = 5 mm, skip = 1 mm, matrix = 256 × 256, with the resolution of three dimensions of one voxel: x = 0.9375 mm, y = 0.9375 mm, z = 6 mm). Next, 96 (144) images per slice were acquired using a gradient echo planar imaging (repetition time = 3,000 or 2,000 msec, matrix = 64 × 64, with the resolution of three dimensions of one voxel: x = 3.75 mm, y = 3.75 mm, z = 6 mm). Finally, a fast spoiled gradient recalled sequence (repetition time = 27 msec, echo time = 6 msec, field of view = 240 × 240 mm, matrix = 256 × 256, with the resolution of three dimensions of one voxel: x = 1.3 mm, y = 0.9375 mm, z = 0.9375 mm) was used in a sagittal plane to collect three-dimensional images covering the entire volume of the brain. The imaging procedures and parameters were similar to those of our previously published studies [6].

Analysis of Functional NeuroImages (AFNI) software (version release in later 2009) was used for data analysis [13]. After correcting for two- and three-dimensional head motion, the functional images were smoothed using an isotropic Gaussian kernel (full width at half maximum = 6 mm). We then used multiple linear regression analysis (using the 3Ddeconvolve program in AFNI) to further correct the head movement artifacts (six estimated motion-induced time series used as non-interest regressors). Finally, a first level fixed-effect statistics was performed using general linear model for each patient at whole-brain level to identify SON-FV-induced BOLD signal increases for generating activation maps.

Due to the inconsistency of spontaneous head movements, we selected a statistical threshold of t >2 (P <0.05, corrected). To avoid false-negative results, a minimum cluster size of 10 voxels was used as an extent threshold. For 13 twice-scanned patients, we separately proceeded data analysis, and then chose the scan with better activation (more volumes or higher t value) as the final result for statistical analysis in the next step.

Since it can be difficult to accurately identify these cortical areas in deformed brains, we deemed normalized analysis to be unsuitable for this cohort of patients with severe brain injury. Thus, after fitting all the patients activation maps individually to their respective structural MRI data, the Heschl gyrus (HG) was defined as the primary auditory cortex [14,15] (if two HG were present, the anterior gyrus was termed area 41 and the posterior gyrus area 42), whilst the planum temporale, the planum polare [16], and the posterior and lateral extensions of HG were defined as the auditory cortices termed area 21/22. Based on these definitions, we chose the bilateral auditory associate cortex in the temporal area as region of interest in each patient after the anatomic landmarks in three orthogonal cross-sectional views [17,18] (axial, coronal, and sagittal) of the individual high-resolution three-dimensional brain images were repeatedly and simultaneously checked by an experienced radiologist (all these steps were carried out using AFNI plugins). Whether the activation of each patient extended to a higher order auditory cortex or not was also determined.

In the case of most VS/UWS patients there was either no activation or activation was found in the primary auditory cortex, which is defined as ‘lower level’ activation. The activation in some VS/UWS patients may extend to a higher order associative auditory cortex (e.g., area 21/22), similar to the activation pattern observed in MCS patients or healthy controls; this type was defined as ‘higher level’ activation [3].

Finally, Receiver Operating Characteristic (ROC) curve analysis was chosen to analyze the prognostic value of primary and secondary auditory cortex (bilateral) activation volume in VS/UWS and MCS patients [19]. Fisher’s exact test looked for differences in outcome depending on the type of cerebral activation. Results were considered significant at P <0.05.