Functional and Structural Deficits in Brain Regions Subserving Face Perception in Schizophrenia

Twenty male patients with chronic schizophrenia and 16 male healthy comparison subjects with normal/corrected vision, age 20–55 years, all right-handed, participated in this study. After a complete description of the study, all participants signed an informed consent form, in accordance with Harvard Medical School and VA Boston Healthcare System guidelines. Demographic data for all subjects are presented in Data Supplement 1, which is available with the online version of this article at http://ajp.psychiatryonline.org. The exclusion criteria were 1) a history of neurological illness or major head trauma that would result in an abnormal EEG, 2) a history of ECT, 3) a history of alcohol or drug dependence, 4) alcohol or drug abuse within the past 5 years, and 5) verbal IQ below 75. Healthy comparison subjects were recruited through newspaper advertisement and were screened with the Structured Clinical Interview for DSM-III-R—Non-Patient Edition (SCID-NP). No comparison subjects had an axis I psychiatric disorder, nor did their first-degree relatives.

The patients were recruited from the VA Boston Healthcare System, Brockton Division, and had a diagnosis of schizophrenia on the basis of their medical records and assessment with the patient edition of the SCID for DSM-IV. All patients were receiving neuroleptic medication, with a mean dose equivalent to 383.8 mg/day of chlorpromazine (SD=266). Five of the 20 patients were receiving typical neuroleptics, 14 were receiving atypical neuroleptics, and one patient was receiving both types of neuroleptics. No significant correlations existed between chlorpromazine-equivalent doses and N170 amplitudes at any electrode site in response to any stimuli (<0.20≤r≤0.24, 0.31≤p≤0.91). The Positive and Negative Syndrome Scale (PANSS) (15) was administered to patients. Mean scores on the PANSS positive symptom scale, negative symptom scale, and the general psychopathological scale were 19.0 (SD=7.7), 21.2 (SD=5.8), and 37.2 (SD=10.0), respectively. Handedness was assessed with the Edinburgh Inventory (16), and the socioeconomic status of subjects and their parents was measured by using the Hollingshead two-factor index.

The EEG was recorded (0.01–100 Hz, 500 Hz digitization) by using 64 silver-silver chloride sintered electrodes in preconfigured caps (ElectroCap International, Eaton, Ohio). The electrode sites were Fp1/Fp2, F7/F8, F5/F6, F3/F4, F1/Fz/F2, FT9/FT10, FT7/FT8, FC5/FC6, FC3/FC4, FC1/FC2, T9/T10, T7/T8, C5/C6, C3/C4, C1/Cz/C2, TP7/TP8, CP5/CP6, CP3/CP4, CP1/CP2, P9/P10, P7/P8, P5/P6, P3/P4, P1/Pz/P2, PO9/PO10, PO7/PO8, PO1/PO2, and O1/Oz/O2 referenced to the right earlobe, with the ground electrode placed on the forehead. The vertical electro-oculogram (EOG) was recorded from two electrodes located medially to the right eye, one above and one below the eye. The horizontal EOG was recorded at the outer canthi. Electrode impedances were <5 kΩ. In the off-line analyses, the event-related potential responses were convolved with a zero phase shift digital low-pass filter at 32 Hz (24 db/octave). Epochs were 800 msec in duration, including a 100-msec prestimulus interval. Epochs at each electrode site were baseline corrected by subtraction of the average prestimulus voltage and mathematically corrected for eye movement artifacts (17). Subsequently, epochs exceeding ±100 μV at any electrode site were rejected, resulting in average of 28.9 trials per stimulus in the patient group and 29.3 trials per stimulus in the comparison group; no group differences were found for any stimulus type (–0.53≤t≤1.0, df=34, 0.31≤p≤0.83).

Final event-related potential responses were recalculated with an average reference montage, consistent with recent event-related potential studies of faces (18, 19). The N170 amplitude peak was defined as the most negative datapoint between 120 and 270 msec poststimulus.

The stimuli were 32 colored pictures of emotionally neutral faces (16 male and 16 female faces; one-half of all faces were Asian, and one-half were Caucasian) (20), 32 pictures of hands, 32 pictures of cars, and 32 pictures of butterflies, all edited to measure 24×24 cm and all viewed by subjects seated 1 m distant from the computer screen. The pseudo-randomly ordered images appeared in the center of the screen for 500 msec, each preceded by a 200-msec interval in which a fixation point appeared in the middle of the screen and then a 200-msec interval in which the screen was blank. The interstimulus interval was 1900 msec. Subjects were instructed to click a mouse button (left and right hands counterbalanced between subjects) as soon as they saw the target picture (butterflies). A no-response was scored if the response time exceeded 1500 msec.

Magnetic resonance images were acquired in 16 of the 20 patients (one patient refused MRI, one had metallic braces, and two were too large for the scanner) and in 13 of 16 healthy comparison subjects (two subjects refused, and one was too large for the scanner). Demographic data for all subjects for whom MRI data were available are presented in Data Supplement 1, which is available with the online version of this article at http://ajp.psychiatryonline.org. The acquisition was done with a 1.5-T General Electric scanner (GE Medical Systems, Milwaukee) at the Brigham and Women’s Hospital in Boston. The mean intervals between EEG and MRI were 10.1 months (SD=9.2, range=0–24) for the healthy comparison subjects and 7.9 months (SD=7.8, range=–3 to 21) for the patients, with no group difference (t=0.71, df=27, p=0.49). The protocol and neuroanatomical boundaries were similar to those used in our previous work (14). Briefly, the anterior landmark was reliably defined by one slice posterior to the mammillary body. The posterior landmark was determined by the anterior tip of the parietal-occipital sulcus in the midsagittal plane. The last slice, including the crux of the fornix, provided the boundary for division of the fusiform gyrus into anterior and posterior subregions. The collateral sulcus was used as the medial border. The occipitotemporal sulcus was used to determine the lateral border. Figure 1 shows the fusiform gyrus in a coronal slice as well as a three-dimensional reconstruction.

Manual drawings of regions of interest were performed on the realigned and resampled coronal slices. The fusiform gyrus of four of the 29 subjects had been delineated in a previous MRI study (14) by the same rater (T.O.). In the remaining 25 subjects, the region was newly delineated for this study. Interrater reliability was computed for the regions of interest by three independent raters who were blind to the subjects’ group membership but not to hemisphere. The intraclass correlations were 0.94 for the left anterior, 0.94 for the right anterior, 0.95 for the left posterior, and 0.95 for the right posterior fusiform gyrus.

Group differences in accuracy in detecting the target pictures and in median reaction times were assessed with t tests. The N170 component was used to assess electrophysiological response to images of faces relative to other stimuli. The event-related potential elicited by the butterfly stimulus, which required a motor response (to ensure that subjects were attending to the stimuli), was not analyzed because we wished to focus on the purely sensory responses. On the basis of previous findings (18, 19), N170 amplitude was measured at P7/P8 and at the more ventral electrode sites P9/P10 and PO9/PO10 (the same level as the inion). The N170 peak amplitude was submitted to repeated-measures analysis of variance (ANOVA) with group as the between-subjects factor and stimulus type (three types), hemisphere (left or right), and electrode site (three sites) as within-subjects factors. The degrees of freedom were adjusted with the Huynh-Feldt epsilon for factors with more than two levels. Two-tailed tests were used, with an alpha of ≤0.05 required for statistical significance.

To determine if some MRI regions of interest were more affected than others, region-of-interest volumes were converted to z scores by using the mean and standard deviation for the comparison group to calculate the z scores, as follows: z_i=(V_i–V_n)/SD_n, where V_i is the individual fusiform gyrus volume and V_n and SD_n are the mean fusiform gyrus volume and standard deviation, respectively, in the comparison group. For the region-of-interest analysis, the standardized scores were submitted to a mixed-model repeated-measures analysis of covariance (ANCOVA) with the total intracranial contents as a covariate, group (schizophrenia patients or comparison subjects) as a between-subjects factor, and hemisphere (left or right) and subregion (anterior or posterior) as within-subjects factors.

As in our previous study (21), Pearson product-moment correlation was used to examine the association between absolute fusiform gyrus gray matter volume and N170 amplitude measured at P7/P8, P9/P10, and PO9/PO10. We hypothesized that smaller fusiform gyrus volumes would be correlated with smaller N170 amplitudes (less negativity) in schizophrenia. Thus, we assessed negative correlations of N170 amplitude at these electrodes with fusiform gyrus volumes (using a significance level of p≤0.05) in both the schizophrenia group and the healthy comparison group.

No significant group differences were found in age, handedness, or parental socioeconomic status. The patients had significantly lower socioeconomic status than the healthy comparison subjects (t=–3.84, df=34, p=0.001), consistent with deficits in functioning due to the disorder. No significant group differences were found in total intracranial contents (t=1.23, df=27, p=0.23), accuracy of target responses (patients: 98.7%; comparison subjects: 99.4%) (t=1.02, df=34, p=0.32), or median reaction time (patients: median=498 msec; comparison subjects: median=478 msec) (t=–0.75, df=34, p=0.46).

For N170 amplitude (illustrated in Figure 2), a repeated-measures ANOVA demonstrated a significant main effect of stimulus (F=4.46, df=2, 33, p=0.02) and electrode site (F=16.03, df=2, 33, p<0.001) but no main effect of hemisphere (F=1.33, df=1, 34, p=0.26) or group (F=3.79, df=1, 34, p=0.06). There was a significant stimulus-by-group interaction (F=4.92, df=2, 33, p=0.02). To further delineate this result, Helmert’s contrasts were performed on the stimulus-by-group interaction. The groups differed in average N170 amplitude in response to images of faces, compared to the average amplitude in response to images of hands and cars (F=9.79, df=1, 34, p=0.004), but they did not differ in the average amplitude in response to images of hands, compared to the average amplitude in response to images of cars (F=2.45, df=1, 34, p=0.13). Thus, the group difference in N170 amplitude was specific to faces. In addition, the N170 amplitude for each stimulus was analyzed by using two-factor ANOVAs, with electrode site (three sites) as a within-subjects factor and group as a between-subjects factor. In these follow-up analyses, the schizophrenia patients showed significant N170 amplitude reduction in response to faces, compared to the healthy comparison subjects (F=6.94, df=1, 34, p=0.01), as well as a reduction in response to images of hands (F=3.41, df=1, 34, p=0.07), but not to images of cars (F=0.75, df=1, 34, p=0.39). Finally, to examine the stimulus effect in each group, N170 amplitude was analyzed using two-factor ANOVAs with stimulus (images of faces, cars, or hands) and electrode site (three sites) as within-subjects factors. The healthy comparison subjects showed a significant main effect of stimulus (F=6.24, df=2, 14, p=0.01), with the most negative N170 amplitude recorded in response to images of faces (group mean: –7.83 μV), relative to images of hands (group mean: –6.44 μV) and cars (group mean: –5.66 μV). The schizophrenia patients did not show this effect (F=0.16, df=2, 18, p=0.86), indicating that their N170 amplitude was not moderated in response to the different stimuli. For the patients, the N170 amplitude was –4.6 μV in response to images of faces, –4.5 μV in response to images of hands, and –4.7 μV in response to images of cars.

The scattergrams in Figure 3 show the anterior and posterior fusiform gyrus absolute volumes on the left and right side in 16 patients with schizophrenia and 13 healthy comparison subjects. The three-factor ANCOVA of the standardized region-of-interest values (z scores) revealed a significant main effect of group (F=6.21, df=1, 27, p=0.02) but no significant main effects of hemisphere (F=2.61, df=1, 27, p=0.12) or subregion (F=0.10, df=1, 27, p=0.76) and no significant group-by-hemisphere-by-subregion (F=0.37, df=1, 48, p=0.55), group-by-hemisphere (F=0.01, df=1, 27, p=0.93), or group-by-subregion interactions (F=0.16, df=1, 27, p=0.70). These results suggest that patients with schizophrenia have bilaterally reduced anterior and posterior fusiform gyrus gray matter volumes, relative to healthy comparison subjects.

For the schizophrenia patients, significant negative correlations were found between right posterior fusiform gyrus volumes and N170 amplitudes in response to images of faces at P8 (r=–0.69, N=16, p=0.003) and PO10 (r=–0.52, N=16, p=0.04) and between left posterior fusiform gyrus volumes and N170 amplitudes in response to images of faces at P8 (r=–0.53, N=16, p=0.03). However, no significant correlations were found between fusiform gyrus volumes and N170 amplitudes in response to images of cars and hands (–0.48≤r≤–0.01, N=16, 0.05<p≤0.97). For the healthy comparison subjects, no significant correlations were found between fusiform gyrus volumes and N170 amplitudes in response to any stimuli (–0.11≤r≤–0.02, N=13, 0.73≤p≤0.95). Table 1 shows the results of Pearson product-moment correlations of absolute fusiform gyrus volumes and N170 amplitudes at P7/P8 in response to images of faces for both groups. The scattergrams in Figure 4 show the relationship between right posterior fusiform gyrus volumes and N170 amplitudes at P8 in response to images of faces in both groups. The correlation coefficients for this relationship were significantly different between groups (z=2.7, p=0.006).

To further explore the specificity of the relationship between N170 amplitudes and fusiform gyrus subregion volumes, we performed multiple linear regression analyses for the correlations that were found to be significant. In each group, the relative volumes of fusiform gyrus subregions were the dependent variables, and the N170 amplitudes in response to images of faces, cars, and hands at each electrode site with a significant correlation of N170 amplitude and fusiform gyrus volume were the independent variables. The results indicated that N170 amplitude at P8 (but not at any other site) in response to images of faces was a significant predictor of right posterior fusiform gyrus volume in the schizophrenia patients (β=–0.18, SE=0.07, t=–2.69, p=0.02) and that N170 amplitudes in response to images of hands and to images of cars were not significant predictors of right posterior fusiform gyrus volume (β=–0.04, SE=0.12, t=–0.37, p=0.72 and β=0.09, SE=0.09, t=0.90, p=0.39, respectively). In addition, N170 amplitudes at P8 in response to any stimuli were not significant predictors of left posterior fusiform gyrus volume in the schizophrenia patients (response to faces: β=–0.04, SE=0.05, t=–0.77, p=0.46; response to hands: β=0.02, SE=0.09, t=0.23, p=0.82; response to cars: β=–0.06, SE=0.07, t=–0.74, p=0.48). We interpreted these results as further evidence supporting the specificity of the relationship between N170 amplitude in response to images of faces, but not other types of stimuli, and volumetric measures of the right posterior fusiform gyrus.