In the present work stroke patients with unilateral left and right hemispheric lesions which frequently included the insula were examined using a facial emotion recognition paradigm while recognition accuracy and reaction time were assessed. In this prospective study we applied both a group comparison between patient groups with left and right hemispheric lesions and healthy controls, and a voxel-based lesion-symptom mapping to investigate the specificity of lesioned brain regions for facial recognition impairment. We enrolled our patients during the acute stage following stroke and asked them to participate in the experiment at a later chronic stage. Together with a comprehensive neuropsychological testing – ensuring that left and right hemispheric lesion groups and matched controls performed comparable for other dimensions such as vision, speech and language, and cognition – we were able to specifically test facial recognition disturbances with a careful control of possibly biasing other impairments.
Exploratory single comparisons suggested that left hemispheric lesions affected recognition performance especially for negative emotional facial expressions (fear and anger) which is in contrast to findings which relate recognition capabilities rather with the right hemisphere (26, 27). Nevertheless, our results are in line with findings by Young et al. (28) who found recognition impairments primarily in patients with left hemispheric lesions following traumatic brain injuries. Actually, fear and anger expressions are difficult to discriminate (29) and impaired recognition for these emotions is typically found in stroke patients (6, 30).
We did not observe impairments in disgust recognition in our patients which is in line with previous findings (7, 31) though also impaired capabilities have been reported (32). In accordance with previous work (6, 21), happiness was the easiest expression to identify compared to negative facial expressions and accurately recognized by our patients. This might be due to the fact that happiness was the only positive expression applied in our study (29). For later studies it might be interesting to use further positive facial expressions (e.g. surprise) to decrease a potential ceiling effect (33).
A more in-depth analysis using VLSM (whole brain), a strategy superior to a gross classification of stroke patients to groups before (34), showed that impaired recognition performances particularly for fear and anger expressions were associated with lesions located predominantly in the left insula. Using ROI analyses (bilateral insular cortex) and correcting for multiple comparisons, a correlation between regions in the left insula and impairments in anger recognition performance was still observable. Our findings support previous work which has revealed similar recognition impairments in patients with lesions of the left insula (14, 33, 35).
Impairment in the recognition of disgust or fear did not stand the correction for multiple comparisons (Table 1). Uncorrected results (Fig. 2) showed highest significance for the recognition of fear in the left medial insula (MNI-coordinates: -34, -1, 15) and for the recognition of disgust for the left ventrolateral prefrontal cortex (vlPFC; MNI-coordinates: -28, 27, 13). The vlPFC area is known to process multidimensional emotional stimuli with respect of the perceived emotional intensity in man (for prosody see 36; for emotional gestures see 37) and monkey (38).
Considering the correction for multiple comparisons, our results corroborate Boucher et al. (33) who found intact disgust recognition in patients with unilateral insular surgery, however they contradict previous findings of impaired disgust recognition in patients with left insular lesions (12, 13). More importantly, our findings question the idea that the insula is primarily relevant for disgust processing (39). Interestingly, our VLSM analysis showed that impairments in recognition of neutral expressions were also associated with lesions in the left insula, hence, corroborating findings by Fusar-Poli et al. (11).
In fact, the insula has a central function in the perception and evaluation of body signals, which is highly critical for the activation of subjective experiences of emotional states but also for the encoding of emotional information displayed for example in facial expressions (40). The anterior and posterior regions of the insula differ not only in terms of their cytoarchitecture, but also in terms of their afferent and efferent connections. While the posterior part of the insular cortex is connected to the thalamus, the somatosensory cortex and the superior temporal gyrus, the anterior insula maintains close connections to the limbic system, particularly to the amygdala, the anterior cingular and prefrontal cortex and the basal ganglia (41). Whereas the posterior insula is responsible for the representation of various visceral functions and physiological body processes (40), the anterior insula is more involved in the evaluation of these body processes, which then forms the basis for various emotional and motivational states. Therefore, according to our data it appears that lesions especially in the left insular cortex cannot that easily be compensated especially for those expressions that are difficult to identify even for healthy participants. In fact, correctly identifying expressions of fear and anger was among the most difficult tasks as shown in a recent work on emotion recognition (42). In contrast, lesions of the right insula might not affect emotion recognition as also previously shown (16).
Limitations
Patients with right hemispheric lesions showed visual search deficits (assessed with the TMT A test). However, since all of the effects for facial emotion recognition were located in the left hemispheric lesion group this did not have an impact on our results. Although post-stroke depression is a common finding (43), self-reported depression severity was comparable between patients and healthy controls with scores reflecting absent or mild symptoms. Thus, depression can be ruled out as a confounding factor biasing social cognition performance in our analyses. Our study comprised a variety of examinations on several days. It therefore cannot be ruled out that especially highly-motivated and well-recovered patients showed willingness for participation, which was critically noted just recently (7). A selection bias could be reflected by relatively small lesion sizes in our sample (mean = 19.9cm³), compared to another recent study (20). This might explain why effects were not that prominent in our study and that impairments were rather selective in contrast for example to findings by Braun et al. (6).