Introduction
Schizophrenia is a chronic and complex mental disorder [
1]. Despite its low prevalence of around 1% in the general population, it accounts for an annual societal cost of more than the annual cost of all cancers combined and the societal financial cost for care is directly linked to the severity of the disease [
2‐
4]. This psychiatric illness is characterized by the presence of positive symptoms (i.e.: delusions, hallucinations) and negative symptoms (i.e.: alogia, avolition, blunted affect, asociality and anhedonia) [
5,
6]. Positive symptoms of schizophrenia are hypothesized to be linked to an increased subcortical release of dopamine, especially in the mesolimbic region (i.e. cortical pathway involving the nucleus accumbens) [
7‐
9]. This results in an increased activity of the dopaminergic receptors D2 and manifests as hallucinations and delusions [
10]. It is hypothesized that the negative symptoms can either be intrinsic to the pathophysiology of schizophrenia or can be secondary symptoms that are related to various factors such as adverse effects of treatment, the environment and comorbidities [
11]. Functional neuroimaging studies also support the evidence of fronto-temporal dysconnectivity in patients suffering from schizophrenia with several frontal lobe and temporal lobe abnormalities that could yield explanations for positive and negative symptoms [
12,
13].
Various antipsychotic pharmaceutical approaches for positive symptoms such as hallucinations are available as first line of treatment [
14,
15]. Anti-dopaminergic medication such as dopamine receptor antagonists (i.e. Risperidone, Quetiapine) and partial dopamine receptor agonists (i.e. Aripiprazole, Brexipiprazole) can be used [
16]. However, around 30% of patients suffering from schizophrenia are said to be treatment resistant as they either fail to respond or only partially respond to two or more antipsychotic medications [
17]. These patients tend to have poorer premorbid social functioning and represent a greater societal financial burden [
18]. For these patients, Clozapine is currently the next line recommended pharmaceutical approach, but up to 60% of the patients on this medication will not respond favourably to treatment [
19‐
21]. For these reasons, various adjunct approaches such as psychological therapies have been developed across the years. The main psychological intervention used for patients with treatment resistant schizophrenia is psychosis oriented cognitive-behavioral therapy (CBT) [
22]. While CBT has been proven effective for reducing positive symptoms for these patients, the results remain sub-optimal and other strategies have been developed to address this limitation [
23,
24].
Amongst these other strategies are virtual reality-based therapies (VRT) such as Avatar Therapy (AT). Developed by Julian Leff and his team in 2008, this psychotherapeutic approach involves the use of an immersive virtual reality system in which patients suffering from treatment resistant schizophrenia (TRS) interact with the Avatar, a virtual representation of their main persistent auditory verbal hallucination which is controlled and animated by the therapist [
25]. Several studies are reporting the effectiveness of AT in the reduction of auditory and verbal hallucinations [
26‐
28]. At the
Institut universitaire en santé mentale de Montréal (IUSMM), AT is a protocolized therapy which is currently being studied with an undergoing trial to compare its effectiveness to CBT. It is designed as a therapeutic process that includes nine therapeutic sessions. The patients attend one AT session per week until completion of the sessions. In the first session the Avatar is being created by the therapist in collaboration with the patient, using a 3D software, to best represent their own representation of their most distressing verbal hallucination. A broad array of features can be employed (gender, facial characteristics, width, height) to design the Avatar. In the remaining eight sessions, the patients will meet and interact with the Avatar using a virtual reality headset. The Avatar is animated by the therapist and the voice of the therapist is modulated using an external voice modifier system to best represent the verbal hallucinations heard by the patient. Facial expressions of the Avatar can be modified in real-time by the therapist by using programmed dimers to modify facial features. The Avatar (including its voice) is therefore personalized for every patient.
While this technique is still being studied and developed, qualitative explorations of the therapeutic processes have been conducted to better understand the intrinsic processes linked to the improvements of patients suffering from TRS undergoing AT. Several themes related to the exchanges between the patient and the Avatar have been elicited and described, as well as the ability to automatically and adequately classify interactions such as self perceptions, beliefs about the voices and emotional responses to them [
29,
30]. However, for the latter, little is known in current scientific literature regarding the expression of emotions by the patient and by the Avatar during the immersive AT sessions.
Emotional expression is crucial to the therapeutic process as it enables empathic abilities [
31]. In addition to emotion attention and clarity, the integration of emotion regulation training to various CBT approaches has been associated with improvement of psychiatric and medical conditions such as persistent physical symptoms and social anxiety [
32,
33].
Despite the blunted affect often portrayed by patients suffering from schizophrenia, they do experience a wide range of emotions; however, clinical access and assessment by the therapist represents a challenging and limiting factor [
34]. Acoustic and vocal cues can be useful tools in the evaluation of expressed emotions [
37]. Vocal cues and variation in audio samples have been studied and employed in the detection of caricatural emotions to assess them in patients with reduced affects or in patients coming from various cultural backgrounds where emotions can be expressed differently [
34,
35].
Considering the importance of identifying emotions in therapeutical processes and their impact on the therapeutic outcome, an exploration of such emotions is needed. The understanding of patient’s emotions as well as the ones expressed by the Avatar in AT to further comprehend the underlying intrinsic therapeutic processes could benefit the outcome of the therapy and ultimately the patient. The aim of this study is to identify the underlying emotions at the core of the patient-Avatar interaction during AT by human-driven qualitative content analysis of immersive sessions transcripts and audio recordings. It is hypothesized that various emotions are experienced throughout the therapeutic process and that those experienced by the patients are often different than those expressed by the Avatar. To our knowledge, no study has yet explored the aspect and dynamic of emotions during AT.
Discussion
The objective of this study was to explore the emotions of patients’ suffering of TRS and that have undergone AT. It was also designed to identify the emotions expressed by the Avatar throughout the immersive sessions. Nine emotions were identified across the transcripts: Anger, Contempt/ Disgust, Fear, Sadness, Shame/ Embarrassment, Interest, Surprise, Joy and Neutral. Neutral, joy and anger were the emotions that were mostly expressed by the patients. As for the Avatar, expression of interest, disgust/contempt and neutral were amongst the emotions the most annotated across the transcripts.
Patients and Avatars in this study expressed various emotions during the psychotherapeutic process of AT. A recent thematic qualitative evaluation of AT involving views of 15 patients on the therapeutic process identified voice embodiment and associated emotions as a major theme, considering that the voice of the Avatar triggers emotional responses [
47]. This can explain why various emotional responses were identified in the presented study and why there seems to be different links between specific verbalizations by the Avatar and the emotional response expressed by the patient. Another recent study explored emotion elicitation in virtual reality for 11 participants and demonstrated the possibility to elicit fear and anger in a secured immersive environment [
48]. Similarly, virtual environments themselves have been found to intrinsically elicit emotions across patients in virtual reality settings [
49].
While similar studies are not identified for patients suffering from TRS, the polarity between anger and joy, the most frequently identified emotions across the transcripts for the patients (after neutral emotions), might be explained by the neurophysiological changes observed in patients suffering from TRS. Neuroimaging studies have reported that emotionally laden images elicited hyper-activation in the dorso-medial prefrontal context and left cerebellum in TRS patients [
50]. In another study, weaker cerebellum activity presented with deficits in emotion recognitions in schizophrenia [
51]. Since the Avatar is visually represented and presents modifications of facial expressions, this might trigger these hyper-activations and oscillations between emotional responses of anger, neutral versus joy, rarely including the other emotions. It is also important to note that emotion identification is possible in patients suffering from schizophrenia and although their affect might not display their emotional processes and responses, they can feel them [
52]. As for the fear response in the patients, it has been found that virtual avatar and human responses can both elicit the same response in the amygdala even if the avatar are overly anthropomorphic [
53]. Further understanding of these emotional responses could be elicited using a wider array of parameters such as heart rate, body temperature, gesture, and overall behavior.
As for the Avatar, as part of the therapeutical processes, it is important for the therapist to elicit patient reactions. Interest, similarly, to most positive emotions, is an emotion that can be strategically employed to create the therapeutic alliance and to reduce the anxiety and fear linked to the therapeutic sessions themselves [
54,
55]. It is also an emotion that can help the therapist to use real and personalized examples to confront or validate the patient, which may explain why interest is such a frequently annotated emotion throughout Avatar transcripts. The emotional expressiveness is an important component of the therapy and it has been demonstrated that simulation of Avatars through virtual reality is a way to train patients suffering from schizophrenia in their abilities to recognize certain emotions [
56].
Limitations
This is an exploratory study and the lack of generalisation of the results can limit its interpretation. Considering the emotional responses, the emotion identification process could have been biased by the coders’ own understanding and perception of emotions. This has been mitigated using an emotional grid that was defined and the interrater analysis. This study does not include the patient’s own labelling and identification of his or her emotions underlying their reactions and transcripts and there could thus be a mismatch between the coders’ perceptions and patients’ perceptions of their own emotions. Another limitation is that visual cues were not taken into consideration which limits the analysis to the content of the verbatims and the audio transcripts. The random selection of the participants could yield to bias considering that emotionally tinged speech can vary for men and women [
57].
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.