Background
Schizophrenia is characterized with neurocognitive and social cognitive impairments, both of which contribute to decreased psychosocial functioning [
1]. Social cognition consists of three major areas: attributional style, theory of mind and emotion recognition [
2]. In the present investigation the focus was on emotion processing and recognition from facial displays.
Brain regions involved in emotion processing are the amygdala, the basal ganglia, the lateral and medial parietal cortex, the medial temporal lobe, the lateral temporal cortex, the dorsal and rostral anterior cingulate cortex, the anterior insular cortex, and the prefrontal cortex [
3,
4]. The amygdala is primarily responsible for the cognitive representation of fear [
5]. The temporal-parietal structures are activated in the process of face and facial expression detection [
6]. The ventromedial prefrontal cortex (VMPFC) and the dorsolateral prefrontal cortex (DLPFC) are involved in the regulation of cognitive functions linked to emotions such as attention, while the ventrolateral prefrontal cortex (VLPFC) signals emotion salience and the need to regulate [
7,
8].The right VLPFC is also implicated in the integration of viscerosensory information with affective signals between the bilateral anterior VLPFCs and the bilateral amygdale [
9] and activated in situations with decision uncertainty [
10]. The anterior insula and anterior cingulate cortex are involved in emotional tasks with cognitive demand [
4].
Multiple previous studies have shown that schizophrenia has an effect on the functioning of the temporal [
11] and prefrontal cortices [
12], including the detection, evaluation and reappraisal of facial emotions. A recent study also revealed that the neural networks of frontolimbic regions are impaired in adolescents at high risk for schizophrenia [
13].
While several previous investigations examined the recognition of basic emotions in schizophrenia [
14], only a few studied the processing of different emotion intensities, while no previous research focused on the evaluation of mixed emotions, however they are requently expressed in everyday situations. Furthermore in many cases negative emotions such as fear are expressed by mixed facial expression (e.g. a mixture of a smile and a fearful facial display). Therefore using mixed emotions can be more ecological valid. Hence in the present investigation facial displays of mixed positive and negative emotions were also presented to patients with schizophrenia and healthy subjects. We hypothesized that the processing of these more complex emotions are more severely impaired in schizophrenia, and therefore a better differentiation of patients from controls can be achieved based on activation patterns. Specifically decreased activations were expected in regions involved in facial emotion processing in the patient group, especially in prefrontal regions, where previous investigation found activation in task with high cognitive demand linked to emotion processing such the anterior insula [
4] and also in regions linked to salience signaling such as the ventrolateral prefrontal cortex [
15].
Discussion
In this study we examined emotion processing in a group of patients with schizophrenia and in healthy controls. The study groups were matched by age, gender, and education. Patients showed a decreased activation in the right anterior insula (RAI), in the right VLPFC, in the right DMPFC during fear, mixed fear, and mixed happy facial processing and showed poorer emotion recognition performance in fear detection relative to controls. In patients on higher doses of APs decreased activations to all emotional displays were detected in the right anterior amygdala, which correlation was the strongest for mixed happy faces.
Subjects were presented with facial displays of happiness, fear, and the mixture of these emotions as described in the methods section. From previous research we know that VLPFC activation is related to emotion intensity and reappraisal [
7,
28], and signals the salience of the stimulus [
8]. Moreover, emotional tasks with cognitive demand involves primarily the anterior cingulate and insula [
4]. Decreased activation was detected in the right anterior insula (RAI), and in the VLPFC region to fear, mixed fear, and mixed happy conditions in patients relative to controls. Previous investigations showed that VLPFC plays a key role in emotion regulation and reappraisal through cortical-subcortical pathways especially to aversive stimuli [
29]. It is also found that VLPFC activity correlated with reduced negative emotional experience during cognitive reappraisal of aversive images [
30]. Taken these results together with our findings raised the possibility that salience signaling and perhaps also emotion regulation linked to negative emotion processing are impaired in schizophrenia, which is in line with previous studies examining emotion processing in schizophrenia that found decreased activation in the VLPFC relative to controls [
12,
15,
31].
Patients showed decreased activation relative to controls in the right middle occipital gyrus (RMOG/BA18) to facial conditions of mixed happy facial stimuli. The RMOG/BA18 is part of the ventral stream of visual processing for which preferential processing of affective stimuli has previously been demonstrated [
32]. Previous studies found increased activation in the RMOG/BA18 to fearful faces compared to happy stimuli during facial emotion processing [
32], furthermore a recent study found that this region is under control by the anterior cingulate cortex during facial emotion processing [
33]. Our finding is in line with previous results showing an early visual processing impairment in emotional facial processing in schizophrenia [
34‐
36]. Furthermore patients with schizophrenia showed decreased activations in the right middle frontal cortex (RMFG/BA10) and in the dorsomedial prefrontal cortex (DMPFC) to fearful facial stimuli. According to previous investigations decreased DMPFC activity in response to emotion facial cues may reflect less cognitive control involved in decoding and/or regulating negative emotions [
37], which was also observed in chronically violent men [
38], and in postpartum depression [
39]. A study examining the theory of mind impairments in subjects with autistic spectrum disorder found decreased activity in the DMPFC during inferring other person’s social emotions in a false belief task [
40].
Patients on higher doses of APs showed decreased activation in the right amygdala to all emotional faces with the strongest correlation for mixed happy condition. Little is known so far on the effects of AP treatment on emotion processing in schizophrenia. In a previous study Sachs et al. [
41] reported no difference in amygdala activation between patients treated with atypical APs and healthy subjects. In another study 12 previously drug-free/naive patients with schizophrenia were treated with olanzapine for 8 weeks and underwent two fMRI scans after 4 and 8 weeks of treatment during implicit and explicit emotional processing. Activity in left amygdala was greater in patients than in controls at the first scan during both explicit and implicit processing, while it was lower in patients at the second relative to the first scan [
42]. However none of the previous studies reported any dose related effect of APs on the activity of brain regions involved in emotion processing. Contrary to previous investigations no between group differences were found in the amygdala activation to fearful faces. A possible explanation to that is that fearful faces in the present study were not as strong fear evoking stimuli as the scenarios depicted in the International Affective Picture Set (IAPS) [
43].
Previous studies found that the superior temporal gyrus and the insular cortex have been involved in the perception of emotions in facial stimuli [
5,
44], while the BA10 region is mainly activated during multi-tasking and theory of mind tasks [
45]. In the present study an increased activation to fearful faces relative to happy faces were detected in the right superior temporal gyrus and in the right middle frontal gyrus (BA10) in healthy subjects, while the opposite difference (happy > fear) was found in patients. In a previous investigation, similarly to our results, an increased activation to fearful stimuli relative to safe stimuli was detected in the insular cortex in healthy subjects [
44].
The activation of the DLPFC is related to the regulation of cognitive processes linked to emotion processing such as attention [
7,
28]. A decreased activation difference between mixed happy faces and happy faces in patients relative to controls were found in the cerebellum (vermis) and in the right DLPFC. Controls showed an increased activation to mixed happy faces relative to happy faces in these regions, while similar between emotion differences were not found in patients. These findings are in line with previous investigations showing that during emotion processing [
46] and social cognition tasks [
47] patients with schizophrenia showed decreased activations in the DLPFC and in the cerebellum during face processing [
48].
Structural impairments in patients may explain the functional differences, since some previous investigations found structural differences in patients with schizophrenia relative to healthy controls in temporal [
49] and frontal regions. However no similar volumetric differences were found in the present study.
A major limitation of the present study was that no neutral facial expressions were used. The reason was that we intended keep the paradigm as short as possible to avoid attention related confounds. A further limitation was that patients were at antipsychotic medication, which may bias the activations, however there were no correlations between antipsychotic doses and BOLD activations in the regions where the between group differences were found. Study groups were matched by age, gender, and education, however this approach has the drawback that it does not take into consideration that patients with schizophrenia experience a significant reduction in educational attainment relative to what we be predicted had they not received such a diagnosis. Due to a technical malfunction of the input device several responses were lost, which made it impossible to include responses into the 1st level model. This may lead to confounds related to the motor activation needed to produce a button press, however no between study group differences were detected in motor areas or supplementary motor regions, which makes it unlikely that this was a major confound.