To determine whether smart conversational agents can be used for detection of neuropsychiatric disorders. Therefore, we reviewed the technologies used, targeted mental disorders and validation ...procedures of relevant proposals in this field.
We searched Scopus, PubMed, Pro-Quest, IEEE Xplore, Web of Science, CINAHL and the Cochrane Library using a predefined search strategy. Studies were included if they focused on neuropsychiatric disorders and involved conversational data for detection and diagnosis. They were assessed for eligibility by independent reviewers and ultimately included if a consensus was reached about their relevance.
2356 references were initially retrieved. Eventually, 17 articles – referring 9 smart conversational agents – met the inclusion criteria. Out of the selected studies, 7 are targeted at neurocognitive disorders, 7 at depression and 3 at other conditions. They apply diverse technological solutions and analysis techniques (82.4% use Artificial Intelligence), and they usually rely on gold standard tests for criterion validity assessment. Acceptability, reliability and other aspects of validity were rarely addressed.
The use of smart conversational agents for the detection of neuropsychiatric disorders is an emerging and promising field of research, with a broad coverage of mental disorders and extended use of AI. However, the few published studies did not undergo robust psychometric validation processes. Future research in this field would benefit from more rigorous validation mechanisms and standardized software and hardware platforms.
Display omitted
•Smart conversational agents can potentially detect neuropsychiatric diseases.•The scientific literature was reviewed using the PRISMA methodology.•The coverage of mental disorders is broad, and the use of AI is widespread.•Emerging and promising field, with lack of robust psychometric validation.
This work explores the advances in conversational agents aimed at the detection of mental health disorders, and specifically the screening of depression. The focus is put on those based on voice ...interaction, but other approaches are also tackled, such as text-based interaction or embodied avatars.
PRISMA was selected as the systematic methodology for the analysis of existing literature, which was retrieved from Scopus, PubMed, IEEE Xplore, APA PsycINFO, Cochrane, and Web of Science. Relevant research addresses the detection of depression using conversational agents, and the selection criteria utilized include their effectiveness, usability, personalization, and psychometric properties.
Of the 993 references initially retrieved, 36 were finally included in our work. The analysis of these studies allowed us to identify 30 conversational agents that claim to detect depression, specifically or in combination with other disorders such as anxiety or stress disorders. As a general approach, screening was implemented in the conversational agents taking as a reference standardized or psychometrically validated clinical tests, which were also utilized as a golden standard for their validation. The implementation of questionnaires such as Patient Health Questionnaire or the Beck Depression Inventory, which are used in 65% of the articles analyzed, stand out.
The usefulness of intelligent conversational agents allows screening to be administered to different types of profiles, such as patients (33% of relevant proposals) and caregivers (11%), although in many cases a target profile is not clearly of (66% of solutions analyzed). This study found 30 standalone conversational agents, but some proposals were explored that combine several approaches for a more enriching data acquisition. The interaction implemented in most relevant conversational agents is text-based, although the evolution is clearly towards voice integration, which in turns enhances their psychometric characteristics, as voice interaction is perceived as more natural and less invasive.
Background:
early detection of dementia and Mild Cognitive Impairment (MCI) have an utmost significance nowadays, and smart conversational agents are becoming more and more capable. DigiMoCA, an ...Alexa-based voice application for the screening of MCI, was developed and tested.
Objective:
to evaluate the acceptability and usability of DigiMoCA, considering the perception of end-users and cognitive assessment administrators, through standard evaluation questionnaires.
Method:
a sample of 46 individuals and 24 evaluators participated in this study. End-users were fairly heterogeneous considering demographic and neuro-psychological characteristics. Evaluators were mostly health and social care professionals, relatively well-balanced in terms of gender, career background and years of experience.
Results:
end-users acceptability ratings were generally positive (rating above 3 in a 5-point scale for all dimensions) and it improved significantly after the interaction with DigiMoCA. Administrators also rated the usability of DigiMoCA, with an average score of 5.86/7 and with high internal consistency (
α
= 0.95).
Conclusion:
although there is still room for improvement in terms of user satisfaction and voice interface, DigiMoCA is perceived as an acceptable, accessible and usable cognitive screening tool, both by individuals being tested and test administrators.
Graphical abstract
Towards a new approach to the Computer Architecture lab Llamas-Nistal, Martin; Fernandez-Iglesias, Manuel J.; Santos-Gago, Juan M. ...
2022 Congreso de Tecnología, Aprendizaje y Enseñanza de la Electrónica (XV Technologies Applied to Electronics Teaching Conference),
2022-June-29
Conference Proceeding
In this paper we present a novel approach to the lab sessions of a undergraduate course on Computer Architectures, based on a real computer built using a Raspberry Pi board. This new approach allows, ...on the one hand, the study of an ARM architecture computer for the CA course; and on the other hand, acquiring basic knowledge related to other disciplines, such as digital electronics and digital systems, thus paving the way for students to take related courses later on.
Short-exercise Assessment in a Computer Architecture Laboratory Llamas-Nistal, MartIn; Liz-DomInguez, MartIn; Santos-Gago, Juan M. ...
2024 XVI Congreso de Tecnología, Aprendizaje y Enseñanza de la Electrónica (TAEE),
2024-June-26
Conference Proceeding
This article introduces a new assessment approach for lab course on Computer Architectures (CA) oof a Telecommunications Technology degree, which can be applicable to other types of lab-bases ...courses. Assessment by means of short exercises in each of the lab sessions is intended to complement classical summative assessment, which was carried out either by means of a final exam or by means of two mid-term exams. This type of assessment introduces an element of formative assessment that allows students to know the degree of understanding of the lab sessions and ultimately to achieve the learning objectives. The results of applying this new approach to the 2022/23 academic year are positive and motivated us to continue to develop it in successive years.
Summary
Predictive tools for major bleeding (MB) using machine learning (ML) might be advantageous over traditional methods. We used data from the Registro Informatizado de Enfermedad TromboEmbólica ...(RIETE) to develop ML algorithms to identify patients with venous thromboembolism (VTE) at increased risk of MB during the first 3 months of anticoagulation. A total of 55 baseline variables were used as predictors. New data prospectively collected from the RIETE were used for further validation. The RIETE and VTE‐BLEED scores were used for comparisons. External validation was performed with the COMMAND‐VTE database. Learning was carried out with data from 49 587 patients, of whom 873 (1.8%) had MB. The best performing ML method was XGBoost. In the prospective validation cohort the sensitivity, specificity, positive predictive value and F1 score were: 33.2%, 93%, 10%, and 15.4% respectively. F1 value for the RIETE and VTE‐BLEED scores were 8.6% and 6.4% respectively. In the external validation cohort the metrics were 10.3%, 87.6%, 3.5% and 5.2% respectively. In that cohort, the F1 value for the RIETE score was 17.3% and for the VTE‐BLEED score 9.75%. The performance of the XGBoost algorithm was better than that from the RIETE and VTE‐BLEED scores only in the prospective validation cohort, but not in the external validation cohort.
PDF
