Diagnosis, treatment, and management of psychiatric disorders is a highly complex task for clinical institutions. Scientific knowledge has advanced significantly since the beginning of psychiatry, leading to standardised approaches to diagnosis and treatment. While these standardised approaches are effective in many cases, the highly individual nature of how psychiatric disorders manifest in different people at different points in time requires the ongoing adaptation of clinical processes to the specific case. Achieving such adaptation is difficult because of the lack of detailed, objective measures of the patient state. MePheSTO is an interdisciplinary research project that envisions a scientifically sound methodology based on artificial intelligence methods for the identification and classification of objective, and thus measurable, digital phenotypes of psychiatric disorders. Important to MePheSTO is the creation of a multimodal corpus including speech, video, and biosensors of social patientclinician interactions, which serves as the basis for deriving methods, models and knowledge. Social interaction, e.g. conversation between a patient and a clinician is traditionally the most important source of information; however, there is a huge challenge in objectively analyzing this in a scalable fashion. Mephesto will support clinicians in this challenging task by objectively measuring and tracking individual patient state by digital as well as interactionbased phenotyping, thereby providing important building blocks towards realising the vision of precision psychiatry. In this symposium, we will present four cardinal use cases in Mephesto that are contributing to this vision and detail the requirements for data recording in each use case.
10:15 Uhr
Quantifying therapeutic alliance within patient-clinician interactions through a multimodal social synchrony model
D. Postin (Bad Zwischenahn , DE)
Details anzeigen
Autor:in:
D. Postin (Bad Zwischenahn , DE)
Therapeutic alliance, i.e. how a patient and a therapist connect, behave, and engage with each other, was shown to be connected to therapy outcome measures robustly across different disorders and therapeutic approaches. However, the psychotherapeutic process is both dynamic and complex. It constitutes a most complex bio-psycho-social system in which language, cognition, and emotions are intertwined and influenced through the interactional dynamics between therapist and patient. The therapeutic alliance can be associated with interpersonal coordination during human interaction in behaviour, physiology, emotional and cognitive states. Numerous studies have made connections to therapeutic processes and outcomes, for instance regarding vocal coordination, body movements, and psychophysiological markers. These markers for therapeutic alliance have the potential to support clinicians during treatment and to assist in increasing the fit between patient and clinician. An important goal of MePheSTO is to develop such digital markers for therapeutic alliance and validate these markers in a multi-lingual, multi-disorder setting. To measure different aspects of social synchrony, video, audio and physiological recordings taken from clinician and patient will be analysed. To validate the measures, clinician- and patient ratings of therapeutic alliance will be utilized alongside therapy outcome measures.
10:37 Uhr
Supporting differential diagnosis of major depressive episode etiology through combined analysis of video, audio and physiology data
A. König (Nice, FR)
Details anzeigen
Autor:innen:
A. König (Nice, FR)
E. Ettore (FR)
P. Müller (DE)
M. Benoit (FR)
M. Benoit (FR)
H. Lindsay (Saarbrücken, DE)
J. Hinze (DE)
D. Postin (DE)
P. Robert (FR)
Background: Major depressive episode (MDE) is a common clinical syndrome. It can be found in different pathologies such as major depressive disorder (MDD), bipolar disorder (BD), post-traumatic stress disorder (PTSD) or even occur in the context of psychological trauma. However, only one syndrome is described in international classifications such as DSM-V, which do not take into account the underlying pathology at the origin of the MDE. Clinical interviews are currently the best source of information to obtain the etiological diagnosis of MDE. Nevertheless, it does not allow an early diagnosis and there are no objective measures of extracted clinical information. To remedy this, the use of digital tools and their correlation with clinical symptomatology seems promising.
Objective: We aim to review the current application of digital tools for MDE diagnosis while highlighting shortcomings for further research. In addition, our work focuses on digital devices easy to use during clinical interview and four most frequently found clinical conditions in which MDE can occur: (1) MDD, (2) BD, (3) PTSD and (4) psychological trauma.
Methods: Most relevant studies in the field were reviewed with a focus on 4 allocated topics of (1) automated voice analysis, (2) behaviour analysis by video and physiological measures, (3) heart rate variability (HRV) and (4) electrodermal activity (EDA).
Results: Based on a narrative review of 74 papers, a digital phenotype of MDE seems to emerge consisting of modifications in speech features (namely temporal, prosodic, spectral, sources, formants and in speech content), modifications in nonverbal behaviour (Head, hand, body and eyes movement,facial expressivity and gaze) and a decrease in physiological measurements (HRV and EDA). We found similarities but also differences when MDE occurs in MDD, BD, PTSD or psychological trauma. However, comparative studies were rare which makes it yet difficult to define a distinct digital phenotype.
10:59 Uhr
Objective measurement of negative and positive symptoms in schizophrenia through automatic speech and language analysis
J. Hinze (Homburg, DE)
Details anzeigen
Autor:in:
J. Hinze (Homburg, DE)
In diagnostic processes and mental status examinations clinicians typically use speech and language as a primary source of information. In schizophrenia-spectrum disorders, speech is used to assess disorganized speech and blunted affect (e.g. intonation). In addition, clinician rated scales (e.g. PANSS) are often used to assess severity of the positive and negative symptoms. However, there is evidence for a disparity in assessing symptom severity between objective speech measures and clinician rated scales (Cohen et al., 2014). In a recent study, a machine learning classifier based on acoustic parameters accurately distinguished not only between healthy participants and patients with an accuracy of 86.2%, but also between patients with predominantly positive vs. negative symptoms with an accuracy of 74.2 (de Boer et al., 2021).
In the Mephesto-project, we plan to analyze audio data of patient-clinician interactions to objectively measure positive and negative symptoms longitudinally in schizophrenia through automatic speech and language analysis. Multiple acoustic parameters, for example the mean of voiced segments per second (reflecting fragmentedness of speech), the standard deviation of the spectral flux (reflecting monotone speech), and the mean and standard deviation of unvoiced segment length (reflecting pause length) will be used to not only differentiate between predominantly positive vs. negative symptoms, but also predict symptom progression in a twelve month interval.
11:21 Uhr
Relapse prediction from longitudinal monitoring based on synchronous remote video interactions
A. König (Nizza, FR)
Details anzeigen
Autor:innen:
A. König (Nizza, FR)
P. Müller (Saarbrücken, DE)
H. Lindsay (Saarbrücken, DE)
J. Hinze (DE)
D. Postin (DE)
A current massive shift towards telemedicine creates unique opportunities but also challenges for patients and clinicians. Especially in psychiatry, a detailed observation of patients’ behaviour during interaction is key to effective and personalised diagnosis and treatment. Such detailed observation is more challenging in telemedicine due to the physical separation between clinician and patient. While recent progress in AI technologies have the potential to be of help in sophisticated interaction analysis, telemedicine systems currently used in psychiatry are limited to rudimentary video call functionalities.
This is due to three main challenges: First, automatic extraction of clinically relevant interaction behaviour is still under-explored in telemedicine. Second, it is not known how such information can be optimally integrated into clinical practice. Third, due to the highly sensitive nature of clinical interactions, ethical and legal considerations have to be satisfied before such data analysis can be realised.
We will present methods and potential system solutions for the support of psychiatric telemedicine interaction addressing these exact key challenges.
The use of different sensor technologies including automated speech, language and video processing can help identify patterns in behaviour by detecting subtle moment-to-moment changes and thus predict eventually risks of relapse. By providing objective and scalable behavior analysis the distance introduced by telemedicine could be compensated for by objective information concerning the progress of the discussion and possibly give indication to the clinician about the necessity of adjusting the course of the interaction. Additional measures in between consultations collected via a mobile application could further supplement clinical diagnosis as well as prognosis for early prevention of aggravation of psychiatric symptoms such as suicidal ideas or drug abuse.