Research article
Yüklə 0,9 Mb. Pdf görüntüsü
|
|
RESEARCH ARTICLE Does Facial Amimia Impact the Recognition of Facial Emotions? An EMG Study in Parkinson ’s Disease Soizic Argaud 1,2 *, Sylvain Delplanque 3 , Jean-François Houvenaghel 1,4 , Manon Auffret 1 , Joan Duprez 1 , Marc Vérin 1,4 , Didier Grandjean 2,3 ‡ , Paul Sauleau 1,5 ‡ 1 Behavior and Basal Ganglia" research unit (EA4712), University of Rennes 1, Rennes, France, 2 Neuroscience of Emotion and Affective Dynamics laboratory, Department of Psychology and Educational Sciences, University of Geneva, Geneva, Switzerland, 3 Swiss Center for Affective Sciences, Campus Biotech, University of Geneva, Geneva, Switzerland, 4 Department of Neurology, Rennes University Hospital, Rennes, France, 5 Department of Neurophysiology, Rennes University Hospital, Rennes, France ‡ These authors are joint senior authors on this work. * argaud.soizic@gmail.com Abstract According to embodied simulation theory, understanding other people ’s emotions is fos- tered by facial mimicry. However, studies assessing the effect of facial mimicry on the rec- ognition of emotion are still controversial. In Parkinson ’s disease (PD), one of the most distinctive clinical features is facial amimia, a reduction in facial expressiveness, but patients also show emotional disturbances. The present study used the pathological model of PD to examine the role of facial mimicry on emotion recognition by investigating EMG responses in PD patients during a facial emotion recognition task (anger, joy, neutral). Our results evidenced a significant decrease in facial mimicry for joy in PD, essentially linked to the absence of reaction of the zygomaticus major and the orbicularis oculi muscles in response to happy avatars, whereas facial mimicry for expressions of anger was relatively preserved. We also confirmed that PD patients were less accurate in recognizing positive and neutral facial expressions and highlighted a beneficial effect of facial mimicry on the recognition of emotion. We thus provide additional arguments for embodied simulation the- ory suggesting that facial mimicry is a potential lever for therapeutic actions in PD even if it seems not to be necessarily required in recognizing emotion as such. Introduction Facial expression is a powerful non-verbal channel providing rapid essential clues for the per- ception of other people ’s emotions, intentions and dispositions during social interactions. It constitutes a key component in daily social communication [ 1 , 2 ]. The processing of facial expressions normally contributes to behaviours that are appropriate to the emotion perceived and the situation and to the person with whom we are communicating. This ensures successful social interactions [ 3 , 4 ]. PLOS ONE | DOI:10.1371/journal.pone.0160329 July 28, 2016 1 / 20
a11111 OPEN ACCESS Citation: Argaud S, Delplanque S, Houvenaghel J-F, Auffret M, Duprez J, Vérin M, et al. (2016) Does Facial Amimia Impact the Recognition of Facial Emotions? An EMG Study in Parkinson ’s Disease. PLoS ONE 11(7): e0160329. doi:10.1371/journal. pone.0160329 Editor: Sonja Kotz, Max Planck Institute for Human Cognitive and Brain Sciences, GERMANY Received: April 13, 2016 Accepted: July 18, 2016 Published: July 28, 2016 Copyright: © 2016 Argaud et al. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: All relevant data are uploaded as Supporting Information files. Funding: This work was supported by the Belgian pharmaceutical company UCB Pharma; the national association of PD patients France Parkinson; and the National Center of Competence in Research “Affective Sciences - Emotions in Individual Behaviour and Social Processes ” (NCCR Affective Sciences) [51NF40-104897 to DG]. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Facial mimicry is defined as the tendency to mimic the facial expressions of individuals with whom we are interacting, or at least to show congruent valence-related facial responses to the perceived expression. According to embodied simulation theory, facial mimicry could foster the understanding of emotion and/or facilitate the inferences and attributions about the mental states of others during social interactions. In the observer, the mirrored expression is linked to a central motor command and is thought to induce a tonic muscular change related to a central proprioceptive feedback; both phenomena are thought to help understand the emotion per- ceived from the clues displayed by the sender [ 5 –
]. In this context, disturbed motor processing can lead to impairments in emotion recognition. This assumption is supported by experimental studies in which the facial feedback was inhib- ited or intensified by behavioural (holding a pen between the lips) or pharmacological (facial botulinum toxin injections) manipulations and by studies among people with facial expression disorders [ 8 –
]. Nevertheless, these studies did not always yield conclusive results. For exam- ple, Bogart and Matsumoto [ 12 ] showed that performances among people with Moebius syn- drome (a congenital condition resulting in facial paralysis) did not differ from those in the control group on a task assessing emotion recognition. The authors concluded that facial mim- icry was not necessarily involved in the process of recognizing emotion. However, people with Moebius syndrome were able to perform as well as healthy people by using compensatory strat- egies, just as they do with their voices and bodies, to convey emotions [ 13 ]. Similarly, correla- tion analyses did not always confirm a relationship between mimicry and emotion recognition among healthy people [ 14 –
]. Parkinson ’s disease (PD) is another pathology affecting facial expression. One of the most distinctive clinical features of this neurodegenerative disorder is facial amimia: the reduction or loss of spontaneous facial movements and emotional facial expressions [ 17 ]. However, PD affects not only facial expressions but also body motion overall and vocal production. Thus, unlike Moebius syndrome people, PD patients may be less prone to compensating for their lack of facial expression by these alternative channels. In addition, PD should not be reduced to motor symptoms since research has shown that it is clearly also characterised by emotional dysfunctions. These disorders concern several components of emotion including subjective feeling, physiological arousal, emotion recognition and motor expression [ 18 ]. Thus, in the light of embodied simulation theory, PD patients may —at least in part—suffer from deficits in emotion recognition as a result of a reduced ability to mimic facial expressions [ 19 – 21 ]. Despite the fact that PD appears to be a useful model to address this issue, there has been no study on this topic to date. Such studies would be valuable given that the few investigations assessing the ability both to express and to recognize facial emotions in PD have evidenced positive correla- tions between impairments in facial expression and disturbances in emotion recognition [ 22
23 ]. In this context, the current study was designed to investigate both the recognition of facial emotion and facial mimicry in PD patients. We expected (1) to confirm the negative impact of PD in the recognition of emotion as reported in the literature, (2) to highlight a facial mimicry disturbance among PD patients and (3) to evidence a link between these two deficits. Materials and Methods Participants Forty patients with Parkinson ’s disease (PD) and 40 healthy controls (HC) took part in the study. All participants provided written informed consents and were informed of the confiden- tial and anonymous aspect of their involvement in this study which was approved by the ethical committee on human experimentation of the Rennes University Hospital, France. All clinical Facial Mimicry in Parkinson's Disease PLOS ONE | DOI:10.1371/journal.pone.0160329 July 28, 2016 2 / 20
Competing Interests: The authors have the following interests: This work was supported in part by the Belgian pharmaceutical company UCB Pharma. There are no patents, products in development or marketed products to declare. This does not alter the authors' adherence to all the PLOS ONE policies on sharing data and materials.
investigation has been conducted according to the principles expressed in the Declaration of Helsinki. Each participant underwent a neuropsychological and psychiatric interview in order to control for any potential bias factors. They were required to obtain a minimum standard score of 5 on the Matrix Reasoning subtest from the Wechsler Adult Intelligence Scale [ 24 ] and
to report normal or corrected-to-normal visual acuity. Concerning visuospatial ability, cut-off scores used for patients inclusion were as follows: 15 on the shape detection screening subtest from the Visual and Object Space Perception battery (VOSP), 18 on the position discrimina- tion VOSP subtest and 7 on the number location VOSP subtest [ 25 ]. Non-emotional face rec- ognition abilities were checked on the Benton unfamiliar-face matching test [ 26 ]; cut-off score for inclusion = 39. No participant had received injections of dermal filling or dermo-contrac- tion agents in the facial muscles; none reported any history of neurological or psychiatric dis- ease (except for PD) or drug/alcohol abuse. To ascertain the absence of apathy, a score higher than -7 on the Lille Apathy Rating Scale (LARS) was required for each patient [ 27 ]. In addition, the participants completed the State-Trait Anxiety Inventory (STAI; [ 28 ]). Disease severity was rated on the Unified Parkinson ’s Disease Rating Scale motor part (UPDRS III; [ 29 ] and the Hoehn and Yahr disability scale [ 30 ], both under dopamine replace- ment therapy (ON DRT) and during a temporary withdrawal from treatment (OFF DRT). For the OFF DRT evaluation, the patients were asked not to take their medication as from the night before the assessment. A levodopa-equivalent daily dose (LEDD) was calculated for each patient [ 31 ]. The patients were on their usual medication during the experiment. Finally, since caring for someone with PD is associated with socio-emotional distress [ 32 , 33 ], caregivers including spouses were not included as HC. Experimental design For each trial, a dynamic avatar appeared on a black screen for 2000 ms ( Fig 1 ). Then, the par- ticipants assessed the emotions portrayed and their intensities on seven visual analogue scales (VAS) labelled joy, sadness, fear, anger, disgust, surprise and neutral. The VAS ranged from 0 to 100%. We used a total of 36 stimuli, presented pseudo-randomly and divided in 3 blocks of 12. In the whole experiment, each participant was exposed to the 36 stimuli with 12 different avatars portrayed 3 different expressions (anger, joy, neutral). The stimuli consisted in natu- rally coloured Caucasian avatars (6 women/6 men). For all stimuli, we used FACSGen [ 34 ,
] to generate videos clips in which the emotional expression unfolded from a neutral state to its Fig 1. Facial emotion recognition task and stimulus material. Timing (A) and response interface (B) of the experiment. Example of facial expression display time (C). The emotional expression unfolded from a neutral state (0) to its emotional peak in 1000 ms (solid line) and stayed at the emotional apex for an additional 1000 ms (dotted line). doi:10.1371/journal.pone.0160329.g001 Facial Mimicry in Parkinson's Disease PLOS ONE | DOI:10.1371/journal.pone.0160329 July 28, 2016 3 / 20
emotional peak in 1000 ms and remained at this level for another 1000 ms. See S1 Appendix for a more detailed description of the experimental material. EMG recordings Bipolar measures of facial EMG activity were performed using Ag/AgCl electrodes, 3 mm in diameter, filled with a highly conductive water electrolyte gel (F-E9M SAFELEAD ™ model and EC60 Signalgel 1 , Grass Technologies). According to the guidelines [ 36 ], six electrodes were positioned over the left zygomaticus (zygomaticus major), orbicularis (orbicularis oculi) and corrugator (corrugator supercilii) muscle regions and one reference electrode (15x20 mm, Ambu 1
pant bias, we recorded facial muscle activity in participants being not aware that this was the real purpose of the electrodes. The use of facial electrodes was explained by the need for record- ing sweat gland activity. The only instruction given to the participants was to assess the emotion portrayed by the avatar using the VAS. No reference to muscle activity or facial movements was given at any time of the experiment. To further confirm that the participants were not aware of the objective of the study, we interviewed them after the experiment. None of the participants reported that they had figured out our purpose or that they had focused on facial movements. The EMG raw signal was recorded with a g.BSamp biosignal amplifier (g.tec), digitized using a 16-bit analog-to-digital converter and stored with a sampling frequency of 1000 Hz. Offline, it was filtered with a 40 –200 Hz band pass filter, rectified and smoothed with a sliding average (window size = 200 ms). PowerLab 16/35 hardware and LabChart 1 Pro Software version 7.3.7 (ADInstrument) were used for EMG data acquisition. After a visual examination of the EMG signals recorded from one second before stimulus onset to stimulus offset, trials in which the EMG amplitude exceeded 30 μV were removed in order to reject any remaining artefacts (3.04% of trials deleted). Data analysis Five trials out of 2880 were excluded from the analyses because no response was recorded on the VAS.
Facial emotion recognition. First, the performances on the emotion recognition task were exploited in terms of emotion decoding accuracy. An expression was considered as accu- rately identified when the emotion receiving the highest intensity rating on the VAS corre- sponded to the emotion displayed (target emotion). The accurately identified expressions were coded as 1; misidentified expressions were coded as 0. Then, in order to determine the nature of the confounding emotions in case of misidentified expressions, confusion percentage was calculated for each non-target emotion; emotions that did not corresponded to the relevant stimulus (number of times each non-target emotion received the highest intensity rating on the VAS instead of the target emotion on the total number of trials). Facial EMG responses. For each trial, the second before stimulus onset was considered as baseline. To examine the temporal profiles of facial reactions, the EMG amplitudes were aver- aged across the sequential 100 ms intervals of stimulus exposure and expressed as a relative percentage of the mean amplitude for baseline. As we expected to highlight a dynamic pattern of facial reactions to emotions as already reported in the literature, facial EMG responses were calculated as previously but on sequential 500 ms periods of stimulus exposure (0 –500; 500– 1000; 1000 –1500 and 1500–2000 ms) to examine the effect of the patients’ clinical characteris- tics as well as to assess the relationship between emotion recognition and facial responses (see S1 Fig
). Facial Mimicry in Parkinson's Disease PLOS ONE | DOI:10.1371/journal.pone.0160329 July 28, 2016 4 / 20
Statistical analysis Data management and statistical analyses were performed using R 3.2.0 [ 37 ]. The significance threshold was set at α = 0.05 except when it was adjusted for multiple comparisons. The experimental fixed effects on decoding accuracy (group and emotion) and on EMG responses (group, emotion, muscle and interval) were tested by fitting logistic and linear mixed models respectively, with random intercepts for both participants and avatars ( “glmer” and “lmer” functions in the “lme4” package, (g)lmer{lme4}; [ 38 ]). In order to control for bias fac- tors, the sociodemographic and neuropsychological variables, as well as their interaction effects with the group factor, were added to these models as fixed factors. Then, analyses of variance (type II Wald Chi-square tests, Anova{car}) were computed and only the potential bias factors with significant effects were retained in the models in order to increase their statistical power. In the case of significant effects, contrasts were tested (testInteractions{phia}) using the Bonfer- roni adjustment method for p values [ 39 ,
]. The fixed effects of group and non-target emotion on confusion percentage were tested similarly by fitting linear mixed models with random intercepts for participants for each level of emotion factor. In order to test whether muscles exhibited higher activity at baseline in the PD patients, analysis of variance was computed on mean EMG amplitudes measured at baseline with group, muscle and emotion as fixed factors and random intercepts for both participants and avatars. As previously, contrasts were tested using the Bonferroni adjustment method for p values in case of significant effects. The impact of medication and other clinical characteristics such as disease severity on decoding accuracy and EMG responses was tested by computing analyses of variance for fitted logistic and linear mixed models with disease duration, the worst affected side, LEDD, Hoehn and Yahr stages and UPDRS III scores (ON and OFF DRT) as fixed factors in addition to the experimental fac- tors (emotion and muscle). For the effects of these clinical characteristics on EMG responses, analyses were conducted on the standardized EMG responses calculated across sequential 500 ms periods of stimulus exposure for each level of muscle, emotion and period factors. Finally, to examine the relationship between emotion recognition and facial mimicry, the fixed effects of the group, the standardized EMG responses of the three recorded muscles calculated on sequential 500 ms periods and their interaction with the group on decoding accuracy were tested by computing analyses of variance for logistic mixed models fitted for each level of emo- tion and period factors. Results
Characteristics of the groups and confounding factors The group characteristics are shown in Table 1 . For details about the patients ’ medication, refer to
S1 Table . All the sociodemographic and neuropsychological factors involving potential bias (age, gender, state and trait anxiety levels, scores on the Matrix and on the Benton tests) were added to the statistical models fitted to assess the effects of experimental factors on decod- ing accuracy and on EMG responses. Only a negative impact of age on decoding accuracy was significant ( χ² = 8.87, df = 1, p = 0.003). Facial emotion decoding accuracy Decoding accuracy varied across experimental conditions ( Fig 2
). The decoding accuracy scores of the PD patients were overall significantly lower than those of the HC ( χ² = 13.34, df = 1, p <0.001). A significant group x emotion interaction effect (χ² = 21.05, df = 2, p<0.001) showed that the difference between groups was interacting with the emotion factor. The decod- ing accuracy scores of the PD patients were significantly lower than those of the HC for happy Facial Mimicry in Parkinson's Disease PLOS ONE | DOI:10.1371/journal.pone.0160329 July 28, 2016 5 / 20
( χ² = 7.07, df = 1, p = 0.024) and neutral avatars (χ² = 14.28, df = 1, p<0.001) but not for angry faces ( χ² = 0.2, df = 1, p = 0.99). In the PD patients, performances tended to increase with the LEDD ( χ² = 2.74, df = 1, p = 0.098). No effect was found for disease duration or severity mea- sured by the Hoehn and Yahr stages or the UPDRS III scores (ON and OFF DRT; all p >0.1).
Table 1. Characteristics of the groups with mean ± standard deviation [range] and statistics. N Healthy controls PD patients (df) F
p value ON DRT
OFF DRT Gender (F/M) 40/40 20/20
20/20 - - - Age (years) 40/40 62.2
± 7.8 [43; 75] 61.2
± 9.6 [42; 79] - (1 ,78) 0.29 0.59
State anxiety [20; 80] 40/38
27.7 ± 7.3 [20; 48] 31.4 ± 6.6 [20; 47] - (1 ,76) 5.61 0.02 Trait anxiety [20; 80] Yüklə 0,9 Mb. Dostları ilə paylaş:
|