Dehbozorgi et al.

BMC Psychiatry (2025) 25:132 BMC Psychiatry

https://doi.org/10.1186/s12888-025-06483-2

SYSTEMATIC REVIEW Open Access

The application of artificial intelligence

in the field of mental health: a systematic review
Raziye Dehbozorgi1,2, Sanaz Zangeneh3, Elham Khooshab4, Donya Hafezi Nia5, Hamid Reza Hanif6,
Pooya Samian7, Mahmoud Yousefi8, Fatemeh Haj Hashemi9, Morteza Vakili10, Neda Jamalimoghadam11* and
Fatemeh Lohrasebi5

Abstract
Introduction The integration of artificial intelligence in mental health care represents a transformative shift
in the identification, treatment, and management of mental disorders. This systematic review explores the diverse
applications of artificial intelligence, emphasizing both its benefits and associated challenges.
Methods A comprehensive literature search was conducted across multiple databases based on Preferred Reporting
Items for Systematic Reviews and Meta-Analyses, including ProQuest, PubMed, Scopus, and Persian databases, result-
ing in 2,638 initial records. After removing duplicates and applying strict selection criteria, 15 articles were included
for analysis.
Results The findings indicate that AI enhances early detection and intervention for mental health conditions. Vari-
ous studies highlighted the effectiveness of AI-driven tools, such as chatbots and predictive modeling, in improving
patient engagement and tailoring interventions. Notably, tools like the Wysa app demonstrated significant improve-
ments in user-reported mental health symptoms. However, ethical considerations regarding data privacy and algo-
rithm transparency emerged as critical challenges.
Discussion While the reviewed studies indicate a generally positive trend in AI applications, some methodologies
exhibited moderate quality, suggesting room for improvement. Involving stakeholders in the creation of AI tech-
nologies is essential for building trust and tackling ethical issues. Future studies should aim to enhance AI methods
and investigate their applicability across various populations.
Conclusion This review underscores the potential of AI to revolutionize mental health care through enhanced acces-
sibility and personalized interventions. However, careful consideration of ethical implications and methodological
rigor is essential to ensure the responsible deployment of AI technologies in this sensitive field.
Keywords Artificial Intelligence, Natural Language Processing, Digital Health, Mental Health, Machine Learning

*Correspondence:
Neda Jamalimoghadam
[email protected]
Full list of author information is available at the end of the article

© The Author(s) 2025. Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0
International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long
as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if
you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or
parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated
otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To
view a copy of this licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by-​nc-​nd/4.​0/.
Dehbozorgi et al. BMC Psychiatry (2025) 25:132 Page 2 of 20

Introduction improve mental health outcomes and reshape the land-

The incorporation of artificial intelligence (AI) into men- scape of mental health care [4].
tal health care has become a revolutionary influence, Despite these promising applications, the integration
altering the ways in which mental health disorders are of AI in mental health care is not without its challenges.
diagnosed, treated, and managed [1]. This systematic Ethical considerations surrounding data privacy and the
review seeks to explore the diverse applications of AI in transparency of AI algorithms are critical issues that
mental health, highlighting both its potential benefits require careful examination. Tavory (2023) highlighted
and the challenges it poses. Over the past few years, the necessity for ethical frameworks to guide the devel-
numerous studies have underscored the capacity of AI opment and deployment of AI technologies in mental
to enhance the early detection and diagnosis of men- health, advocating for a balanced approach that prior-
tal health conditions, facilitating timely and appropriate itizes patient safety and informed consent [5].
interventions. Tornero-Costa et al. (2022) critically assess the meth-
The application of AI in mental health care not only odological and quality flaws present in the use of AI
enhances early detection and diagnosis but also enables within mental health research. The authors identify sev-
personalized treatment and improved patient engage- eral key issues that undermine the reliability and validity
ment. AI analyzes patient data to tailor interventions of AI applications in this field, including inadequate study
and includes tools like chatbots for support, especially designs, insufficient sample sizes, and a lack of trans-
in areas with limited access to traditional services. How- parency in reporting methodologies. Furthermore, they
ever, challenges such as ethical issues, data privacy, and emphasize the importance of rigorous methodological
the need for effective implementation frameworks must standards to enhance the credibility of AI-driven research
be addressed. This review will explore these aspects to outcomes. By highlighting these flaws, the review under-
ensure responsible integration of AI into mental health scores the need for improved research practices to ensure
systems [2]. that AI technologies can be effectively and ethically inte-
Gaffney and colleagues worked on the AI in mental grated into mental health care, ultimately advancing the
health care in their article. They discuss the significant role field and benefiting patient outcomes [6].
of conversational agents, which are AI-driven tools such as Furthermore, the previous review noted critical issues
chatbots, in providing immediate support and interven- such as inadequate transparency in reporting AI model
tion for individuals experiencing mental health issues. The features and data preprocessing techniques, which are
review highlights that conversational agents can enhance essential for ensuring reproducibility and reliability in
accessibility to mental health resources, deliver person- research. Many studies included in the earlier review did
alized therapeutic interactions, and facilitate ongoing not sufficiently assess the quality of data or the appropri-
engagement with users. By leveraging natural language ateness of AI methodologies for their specific applica-
processing and machine learning, these agents can effec- tions, leading to potential biases and an overly optimistic
tively assess user needs, provide tailored responses, and view of AI performance. These shortcomings underscore
promote self-management strategies, ultimately contribut- the necessity for a new systematic review that not only
ing to improved mental health outcomes [3]. addresses these methodological gaps but also provides
In an examination of the transformative potential of AI a detailed overview of AI applications in mental health,
in mental health care, Olawade et al. (2022) emphasize including their benefits and challenges.
the increasing integration of AI technologies into mental This article seeks to address this gap by consolidating
health services, which offer innovative solutions for diag- recent research findings, assessing methodologies, and
nosis, treatment, and patient engagement. Current trends pinpointing areas for further exploration. Through this
suggest that AI can facilitate the early detection of men- approach, we aim to enhance the understanding of how
tal health disorders, optimize therapeutic interventions AI can be successfully incorporated into mental health
through personalized approaches, and enhance access care, ultimately improving the quality of care for those
to care, particularly for underserved populations. Addi- experiencing mental health difficulties.
tionally, the authors discuss future prospects, highlight-
ing the importance of ethical considerations and robust
implementation strategies to ensure the effective integra- Methods
tion of AI applications into clinical practice. By address- Reviewing the literature
ing both the opportunities and challenges associated with The systematic review on the application of AI in mental
AI in mental health, this review aims to provide a com- health involved a comprehensive search across multiple
prehensive understanding of how these technologies can databases to gather a robust collection of relevant studies.
Dehbozorgi et al. BMC Psychiatry (2025) 25:132 Page 3 of 20

The search encompassed both Persian databases, to the identification of 369 articles pertinent to the topic
including IranDoc and ISC, and international databases by July 15, 2024. We then examined the abstracts of the
such as ProQuest, PubMed, Scopus, Web of Science, remaining articles, yielding 80 articles that met the ini-
Cochrane, Wiley, and the American Psychiatric Associa- tial criteria. However, 23 articles were excluded due to
tion’s APA database. the lack of full text, reducing the number to 57 from July
For the initial search phase, we employed a search 16–29, 2024. A thorough review of the full-text articles
syntax focusing on the terms “artificial intelligence” and resulted in the exclusion of 42 articles based on specific
“mental health.” The results varied significantly, with criteria, leaving 15 articles for final inclusion from August
international databases yielding a higher number of 2–15, 2024. The reasons for excluding articles at each
records. Specifically, ProQuest identified 944 records, stage of the systematic review process are as follows: Ini-
PubMed found 829, and Scopus retrieved 353.In contrast, tially, from the 2,638 records identified, 1,167 duplicates
Persian databases like IranDoc and ISC produced no rel- were removed, resulting in 1,471 unique articles. Dur-
evant records, indicating a limited volume of research ing the title screening phase, 1,471 articles were assessed
in the application of AI in mental health. Notably, SID, for relevance, leading to the exclusion of 1,102 articles
a Persian database, contributed four records, while Web that did not meet the criteria based on their titles. Sub-
of Science retrieved 171 records, and E-journal plat- sequently, the remaining 369 articles were further evalu-
forms like Cochrane and Wiley yielded 65 and 7 records, ated through abstract screening, where 289 articles were
respectively. Overall, the search across all databases gen- excluded due to insufficient relevance to the research
erated 2,638 initial records. question or lack of focus on artificial intelligence applica-
To refine the dataset, we removed duplicate entries, tions in mental health. In the full-text review phase, 57
reducing the total to 1,471 unique records. Following articles were considered, but 42 were excluded for rea-
this, we conducted a title screening phase to further sons such as the absence of full-text availability, lack of
narrow down the records to those most relevant to our methodological rigor, or failure to address the specific
research question. After this rigorous screening, only 369 aims of the review. This left 15 articles that met all inclu-
records were deemed suitable for the subsequent phases sion criteria. Each of these articles underwent a compre-
of the systematic review process. hensive risk of bias assessment, ensuring that the final
This comprehensive approach highlights the diversity of selection was based on robust methodological standards
resources and the necessity of detailed screening processes and relevance to the topic of AI in mental health Table 1.
in conducting a systematic review, particularly when ana- We conducted a comprehensive risk of bias assessment
lyzing the intersection of AI and mental health research. for the 15 final articles to evaluate their methodological
This comprehensive approach highlights the diver- rigor and reliability, concluding this assessment phase
sity of resources and the necessity of detailed screening on August 29, 2024. Data extraction from the selected
processes in conducting a systematic review, particularly studies was completed between September 5–15, 2024,
when analyzing the intersection of AI and mental health ensuring that each study’s findings and details were sys-
research.The search strategy, along with thorough dedu- tematically recorded for further analysis. This rigorous
plication and screening, allowed for a focused dataset multi-stage approach facilitated a focused and high-qual-
essential for synthesizing accurate insights into AI’s role ity selection of studies that addressed the application of
and effectiveness within the mental health domain. artificial intelligence in mental health (see Fig. 1).

Selection criteria and search strategy

The systematic review followed a multi-stage selection Quality appraisal methods
process to ensure a thorough and unbiased analysis. Ini- The comparative analysis of the 15 studies reveals sig-
tially, we retrieved a total of 2, 638 records from various nificant insights regarding the quality assessment
databases, including ProQuest, PubMed, Scopus, Web of tools utilized in AI applications within mental health
Science, Cochrane, and Wiley, along with Persian data- interventions.
bases such as IranDoc, ISC, and SID. Searches were con- The studies, spanning from 2018 to 2023, illustrate a
ducted using specific syntaxes tailored to each database, growing interest in the integration of AI technologies in
primarily combining the terms “artificial intelligence” mental health research. Various quality assessment tools
and “mental health” to capture relevant literature. were used across these studies, reflecting the researchers’
From the initial 2,638 records, we identified and commitment to rigorous evaluation of their methodolo-
removed 1,167 duplicates, resulting in a pool of 1,471 gies (see Table 1).
articles as of July 12, 2024. In the first screening stage, we A notable number of studies utilized the Mixed Meth-
reviewed these articles based on title relevance, leading ods Appraisal Tool (MMAT), which was applied in four
Dehbozorgi et al. BMC Psychiatry (2025) 25:132 Page 4 of 20

studies, including those by Inkster et al. (2018), Chin Lei (2023) and Xiao Li (2023) explored more advanced
et al. (2021), Götzl et al. (2022), and Pei (2022). The AI techniques, with Lei et al. using Long Short-Term
MMAT is particularly effective for evaluating mixed- Memory (LSTM) models for emotion recognition tasks
methods research, allowing for a nuanced understand- and Li applying deep learning (CNN) to assess mental
ing of complex mental health dynamics. Its application health states by analyzing social media text data. These
across multiple studies underscores its robustness in approaches underscore the adaptability of AI in process-
assessing research quality in this field [7]. ing complex data patterns, such as emotional and mental
The Newcastle–Ottawa Scale (NOS) was another health indicators [12, 27].
prominent tool, used in several studies, including those Götzl et al. (2022) summarized AI’s broader purpose
by Xiao Li (2023), Pei (2022), and Tate (2020). The NOS is as learning systems that evaluate large information sets,
well-regarded for its application in observational studies, often applied in platforms like YouTube and Spotify,
providing a systematic approach to evaluating the qual- showcasing AI’s versatility in various interactive media
ity of non-randomized studies. The consistent use of the [28].
NOS across these studies indicates its reliability and rel- These studies reveal the diversity and scope of AI appli-
evance in assessing the quality of observational research cations in mental health, ranging from emotional rec-
in AI-driven mental health interventions [8]. ognition and chatbot support to large-scale predictive
Additionally, the Joanna Briggs Institute (JBI) tool models, illustrating AI’s role in advancing personalized
was employed in various experimental studies, includ- and accessible mental health solutions.
ing those by Didarul Alam et al. (2021), Yanqi Guo
(2023), Ilona Halim et al. (2023), and Naveen Kumari & Tools and algorithms used in AI in mental health
Rekha Bhatia (2022). The JBI tool is designed to assess The comparative analysis of the 15 articles concerning
the methodological quality of experimental and quasi- the application of AI in mental health reveals a diverse
experimental studies, contributing to a comprehensive landscape of methodologies, participant demograph-
understanding of the effectiveness of AI interventions in ics, and AI technologies. Each study contributes unique
mental health [9]. insights into the effectiveness and acceptability of AI-
Despite the predominance of established tools like the driven interventions (Table 2).
MMAT, NOS, and JBI, variability in the quality of stud- The first study by Inkster et al. (2018) employed a
ies was observed, particularly with the application of dif- mixed-methods approach to evaluate the Wysa app,
ferent methodologies. For instance, the study by Rohit an AI-enabled chatbot designed for mental well-being.
Rastogi et al. (2022) utilized the JBI tool but received a This study focused on a global user base of individu-
moderate quality score, indicating that even established als self-reporting symptoms of depression. The find-
assessment tools can yield varied results based on the ings indicated that users who engaged more frequently
study design and execution. with the app experienced significant improvements in
In summary, the analysis of the quality assessment tools their depressive symptoms, with 67.7% of participants
used in these studies highlights the importance of meth- finding the app helpful. This highlights the potential
odological rigor in AI-related mental health research. The effectiveness of text-based conversational AI in mental
diversity of tools employed—MMAT, NOS, and JBI— health support [10].
demonstrates the multifaceted nature of research meth- In contrast, Xiao Li (2023) conducted a quantitative
odologies in this area. The consistent application of these study aimed at understanding depression among the
established tools enhances the credibility of findings and elderly. Utilizing Long Short-Term Memory (LSTM)
paves the way for future advancements in the integration networks, the research focused on emotional recogni-
of AI technologies in mental health interventions. tion, demonstrating how AI can model temporal infor-
mation to improve mental health interventions. The
integration of AI attention mechanisms enhanced the
Results accuracy of emotional feature recognition, suggest-
Application of AI in mental health ing that advanced AI techniques can significantly con-
The 15 reviewed studies illustrate various interpretations tribute to understanding and addressing mental health
and applications of AI in mental health contexts. Ink- issues [12].
ster et al. (2018) introduced Wysa, an AI-driven chatbot Liang et al. (2022) evaluated the mental health of col-
aimed at improving mental well-being by enabling self- lege students using a Convolutional Neural Network
reflection and resilience through conversational support (CNN) to analyze text data from online forums. This
[10](Table 2). experimental study indicated that AI could effectively
Dehbozorgi et al. BMC Psychiatry (2025) 25:132 Page 5 of 20

Table 1 Search strategy (search syntax: (artificial intelligence AND mental))

Databases Number of records Number of records after deleting After title
duplications sceening

IranDoc(Persian) 0 0 0
ISC(Persian) 0 0 0
proquest 100 91 91
Pubmed 829 794
scopus 353 351 53
SID(Persian) 4 1 1
Web Of Science 171 167 167
E Journals Number of records Search syntax
Cochrane 65 45 45
Wiley 7 1 1
Magiran(Persian) 0 0 0
APA: American Psychiatric Association 265 12 11
Total 2638 1471 369

monitor and assess mental health states, providing per- Tate (2020) created a model employing machine learn-
sonalized psychological support based on user-generated ing techniques to forecast mental health issues in adoles-
content. The study’s results affirmed the feasibility of AI cents, attaining a satisfactory level of predictive accuracy.
in delivering tailored mental health interventions [13]. While the model was not appropriate for clinical applica-
Chin et al. (2023) explored user interactions with the tion, it established a foundation for future investigations
SimSimi chatbot across various cultural contexts, ana- into predicting mental health in adolescents [31].
lyzing 96,197 conversations from Eastern and West- Didarul Alam et al. (2021) explored the adoption
ern countries. This mixed-methods study revealed of mobile health (mHealth) technologies during the
cultural differences in the expression of depressive COVID-19 pandemic by utilizing Structural Equation
moods, emphasizing the chatbot’s role in social science Modeling and Artificial Neural Networks (ANN). The
research and its high acceptability among users. The find- results demonstrated a positive correlation between
ings suggest that AI can facilitate large-scale data collec- mHealth usage and mental well-being, emphasizing the
tion and cultural analysis of mental health expressions role of AI in improving user engagement during times of
[29]. crisis [19].
Dadi (2021) utilized machine learning to analyze mul- Götzl et al. (2022) explored young people’s attitudes
timodal data from the UK Biobank, demonstrating how toward AI-informed mHealth apps, revealing a generally
population modeling can derive mental health meas- positive reception and willingness to use AI for personal-
ures from diverse inputs, including brain imaging. This ized interventions. This study emphasizes the potential of
approach indicates that AI can complement traditional AI to adapt interventions to individual needs, enhancing
psychometric assessments, enhancing the understanding the effectiveness of mental health support [28].
of mental health determinants in large populations [15]. Onuki et al. (2022) examined the use of wearable
Rathnayaka (2022) focused on the design and devel- devices to estimate mental health conditions based on
opment of a BA-based AI chatbot, confirming its effec- heart rate data, finding high acceptability among users.
tiveness in providing emotional support and remote This study illustrates how AI can integrate with wearable
mental health monitoring. The participatory evalua- technology to provide real-time health monitoring [21].
tion highlighted user feedback on the chatbot’s capabili- Alamgir et al. (2023) focused on facial expression rec-
ties, reinforcing the importance of user experience in AI ognition using a hybrid AI model, demonstrating its
applications for mental health [30]. ability to categorize emotions effectively. This research
Pei (2022) utilized a neural network algorithm to fore- highlights the potential of AI in enhancing emotional
cast and assess the mental health status of college stu- awareness and understanding in mental health contexts
dents, uncovering notable psychological stressors. This [32].
research highlights the effectiveness of AI in recogniz- Yanqi Guo (2023) analyzed the impact of AI technology
ing mental health risks and customizing interventions on family education, revealing significant improvements
through predictive analytics [17]. in mental health outcomes among students receiving
Dehbozorgi et al. BMC Psychiatry (2025) 25:132 Page 6 of 20

Fig. 1 Diagram for the inclusion and exclusion of articles

AI-based support compared to traditional methods. This mental health care by offering personalized, accessible,
study showcases the versatility of AI applications beyond and effective support. As the field progresses, additional
direct mental health interventions [23]. research is crucial to enhance these technologies and
Finally, Halim et al. (2023) tested an individualized vir- confirm their effectiveness across diverse populations
tual reality (iVR) approach to improve self-compassion (Table 3).
and reduce depressive symptoms. Although primar-
ily focused on VR, the study emphasized the potential Quality appraisal methods
of integrating AI to personalize user experiences and In examining the application of AI in mental health stud-
enhance therapeutic outcomes [33]. ies, various study types and quality assessment tools were
In conclusion, the combined results from these stud- employed, with each study exhibiting unique methodo-
ies reflect an increasing acceptance and effectiveness of logical strengths (Table 1).
AI in mental health interventions across different demo- Inkster et al. (2018) [10] conducted a mixed-methods
graphics and methodologies. The varied applications of study, receiving a total score of 7, assessed as “good” qual-
AI, including chatbots, predictive models, and wearable ity using the Mixed-Methods Appraisal Tool (MMAT),
technologies, underscore its potential to revolutionize demonstrating effective integration of both qualitative
Dehbozorgi et al. BMC Psychiatry (2025) 25:132 Page 7 of 20

Table 2 Quality assessment of articles

Row Author Year Study Quality Total score Study quality Reference
type assessment
tools

1 Inkster B etal 2018 Mixed-Methods Study MMAT 7 good [11]

2 Xiao Li 2023 Quantitative NOS 7 good [12]
3 Liang L, 2022 Experimental study NOS 5 good [13]
4 Chin et al., 2023 Mixed Methods Study MMAT 7 good [14]
5 Dadi 2021 Population modeling NOS 6 moderate [15]
6 Rathnayaka 2022 designing and developing of BA-based AI NOS 6 moderate [16]
chatbot and a pilot study
7 Pei 2022 Two phase ( Prediction and Analysis) CASP 7 good [17]
8 Tate 2020 Model development NOS 7 good [18]
9 Didarul Alam et. al −2021 empirical study with JBI 10 good [19]
SEMANN
(Structural Equation Modeling – Artificial
Neural Network
10 Götzl et al −2022 parallel mixed–method study MMAT 7 good [20]
11 Onuki. M 2022 Casual- NOS 6 Moderate [21]
Et al comparative
research
12 Alamgir FM et al 2022 Multi-methods NOS 7 good [22]
study
13 Yanqi Guo, 2023 Experimental study (Randomized controlled JBI 9 Good [23]
trials)
14 Ilona Halim, et al 2023 Experimental study (Randomized controlled JBI 7 Good [24]
trials)
15 Naveen Kumari· Rekha Bhatia 2022 experimental JBI 8 Good [25]
16 Rohit Rastogi, etal 2022 experimental JBI 5 Moderate [26]

and quantitative data. Similarly, Xiao Li (2023) [12] SEMANN (Structural Equation Modeling – Artificial
and Chin et al. (2023) [29] employed quantitative and Neural Network), both achieving “good” quality with
mixed-methods approaches, respectively, with their stud- a score of 7 under NOS. Didarul Alam et al. (2021) [19]
ies scored at 7 by the Newcastle–Ottawa Scale (NOS), conducted a parallel mixed-method study, assessed by
indicating high methodological quality and thorough the Joanna Briggs Institute (JBI) checklist, scoring 10 and
evaluation of mental health indicators related to AI rated “good,” reflecting strong methodological coherence
applications. across study phases.
Dadi (2021) [15] focused on population modeling with In randomized controlled trials, Götzl et al. (2022) [28]
AI, achieving a “good” quality rating and a score of 6 from and Onuki et al. (2022) [21] both received “good” ratings
NOS. In a similar context, Liang et al. (2022) [13] con- with scores of 9 and 7, respectively, indicating robust
ducted an experimental study focused on AI model per- experimental control and rigorous assessment in AI
formance, assessed by NOS and rated “good” with a score application to mental health. Similarly, Alamgir FM et al.
of 5, reflecting methodological rigor but with room for (2022) [32] and Yanqi Guo et al. (2023) [23] conducted
further improvement. experimental studies, each with “good” quality ratings,
A more complex study design was employed by Rath- showing consistency in applying AI within structured
nayaka (2022) [30], who utilized the two-phase predic- trials.
tion and analysis approach to AI applications. This study Finally, Kumari et al. (2023) [25] and Rohit Rastogi
received a score of 6 on NOS, indicating “moderate” qual- et al. (2022) [34] utilized a more experimental frame-
ity, suggesting effective use of predictive techniques but work, rated as “moderate” quality with scores of 8 and
with potential limitations in methodological consistency. 5 under JBI, illustrating potential for reliable AI-driven
Conversely, Pei (2022) [17] and Tate (2020) [31] under- results but highlighting the need for additional methodo-
took model development and empirical study using logical refinement.
Table 3 Extracted data from articles
N Author Study type Study aims participants methodology analysis Meaning of Artificial Acceptability of application of Effectiveness of Artificial Kind of Name of Artificial
and year Intelligence Artificial Intelligence Artificial Intelligence Intelligence Artificial Intelligence
Intelligence

1 Inkster B Mixed-Methods This study aimed a group of anonymous In the study, a group The quantitative analy- Wysa, a Smartphone- The real-world data determine the effec- The average mood improvement Chatbot App Wysa app:
et al Study to present a pre- global users were of anonymous global users sis measured the app Based Empathetic evaluation findings tiveness of delivering (ie, difference in pre- and post-
2018 [10, liminary real-world observed who voluntarily were observed who voluntarily impact by comparing Artificial on the effectiveness positive psychology self-reported depression scores)
11] data evaluation installed the Wysa app, installed the Wysa app, engaged the average improve- Intelligence Chatbot and engagement levels and mental well-being between the groups (ie, high vs low
of the effectiveness engaged in text-based intext-based messaging, ment in symptoms App for Mental Well- of Wysa app on users techniques in a text- users; n = 108 and n = 21, respectively)
and engagement messaging, and self- and self-reported symptoms of depression Being with self-reported based conversational revealed that the high users group
levels of an AI- reported symptoms of depression using the Patient between high and low Wysa, developed symptoms of depres- mode using the Wysa had significantly higher average
Dehbozorgi et al. BMC Psychiatry

enabled, empa- of depression using Health Questionnaire-9. On users. The qualitative by Touchkin, is an AI- sion show promise app on users with self- improvement (mean 5.84 [SD 6.66])
thetic, text-based the Patient Health the basis of theextent of app analysis measured based emotionally reported symptoms compared with the low users group
conversational Questionnaire-9 usage on and between 2 con- the app engagement intelligent mobile chat- of depression. Users (mean 3.52 [SD 6.15]); Mann–Whitney
mobile mental well- 2 distinct groups of users secutive screening time points, and experience by ana- bot app aimed at build- were presented P = .03 and with a moderate effect
being app, Wysa, (high users and low users) 2 distinct groups of users (high lyzing in-app user ing mental resilience with the validated size of 0.63. Moreover, 67.7%
on users with self- emerged users and low users) emerged. feedback and evaluated and promoting mental Patient Health of user-provided feedback responses
reported symptoms The study used mixed-methods the performance well-being using a text- Questionnaire (PHQ-9) found the app experience helpful
of depression approach to evaluate the impact of a machine learning based conversational during their conversa- and encouraging
and engagement levels classifier to detect user interface. The Wysa app tions and screened
(2025) 25:132

among these users. Thequantita- objections during con- assists users to develop for selection based
tive analysis measured the app versations positive self-expression on their 2-item (PHQ-2)
impact by comparing the aver- by using AI to create score. The average
age improvement in symptoms an external and respon- improvement in self-
of depression between highand sive self-reflection reported symptoms
low users. The qualitative analysis environment of depression
measured the app engagement (Pre-PHQ-9 minus Post-
and experience by analyzing in- PHQ-9) was compared
app user feedback andevaluated between 2 comparison
the performance of a machine groups: (1) more
learning classifier to detect user engaged app users
objections during conversations (“high users” group)
and (2) less engaged
app users (“low users”
group)
Improvement
of depressive
symptoms
2 Xiao Li Quantitative The purpose elderly in the two In order to make the output Quantitative methods, AI, specifically LSTM, AI attention mecha- AI was applied The study integrated AI attention LSTM with AI LSTM with AI atten-
2023 [12] is to understand groups of community of different emotional data including t-tests, was used to model nisms were explored through AI atten- mechanisms with LSTM, significantly attention tion mechanisms
the depression ( control group (n = 46) in LSTM more discriminative, variance analysis, temporal information to enhance LSTM tion in both time improving emotion recognition mechanisms focused on enhanc-
status of the elderly and the intervention a method to dynamically filter and multiple stepwise crucial for emotion outputs, improving and feature dimensions performance compared to traditional focused ing emotional
in the commu- group (n = 47)) the output of LSTM is proposed. regression, were used recognition tasks emotional feature to optimize LSTM methods on enhanc- feature recognition
nity, explore its Combining the methods for data analysis recognition and clas- outputs for speech ing emo- in speech
influencing factors, of Attention-LSTM, time- sification accuracy emotion recognition tional feature
formulate a com- dimensional AI recognition
prehensive psycho- attention, and feature-dimen- in speech
logical intervention sional AI attention, the best
plan according model in this paper is obtained
to the influencing
factors, implement
demonstration
psychological nter-
vention, and evalu-
ate and feedback
the effect, so as to
provide a reference
for improving
the mental health
of the elderly
Page 8 of 20
Table 3 (continued)
N Author Study type Study aims participants methodology analysis Meaning of Artificial Acceptability of application of Effectiveness of Artificial Kind of Name of Artificial
and year Intelligence Artificial Intelligence Artificial Intelligence Intelligence Artificial Intelligence
Intelligence

3 Liang L, Experimental To evaluate 185 students in sports □ Data Collection: □ CNN Model: The AI, particularly deep The platform AI is applied to evaluate The CNN model showed improved Deep Not explicitly
2022 [13] study the effectiveness major Mental health data were CNN was trained learning through CNN, has achieved good and provide early performance over traditional Learning, named, but involves
of artificial intelli- ( Training set: 137 collected from internal forums and tested using forum is used to monitor results. The mental warnings regard- models (e.g., FastText) in recognizing specifically CNN as the core AI
gence (AI) in assess- Test set: 48) and through the Symptom data, with specific and assess the mental health plan for college ing the mental health and classifying mental health states, Convolu- technology
ing and improving Checklist-90 (SCL-90) question- parameters set to opti- health state of college students is feasible of students, and to offer with higher accuracy and F1 scores tional Neural
the mental health naire mize its performance students by analyzing personalized Networks
of college students □ AI Model: A Convolu- □ Questionnaire text data from forums psychological support (CNN)
tional Neural Network (CNN) Data: The SCL-90 based on the analysis
was employed to process results were statistically of forum posts
Dehbozorgi et al. BMC Psychiatry

and analyze the text data analyzed to compare

from the forums mental health states
□ Questionnaire Design: The under traditional
SCL-90 was used to evaluate and AI-enhanced
the mental health status education
of the students
4 Chin et al., Mixed Methods To analyze user SimSimi Chatbot Analyzed user-generated Data categorized by cul- SimSimi is an open- Widely used across 111 Used to study Effective in handling diverse user SimSimi SimSimi Chatbot
2023 [14, Study interactions users. Data collected conversations with SimSimi tural groups (Eastern vs. domain chatbot languages with 400 cultural differences inputs and generating appropriate operates
29] with SimSimi from 96,197 conversa- from May 2016 to December Western) to compare designed for social million users, indicating in how users express responses, facilitating large-scale data as a conver-
(2025) 25:132

chatbot regard- tions in Eastern countries 2020, focusing on expressions frequency and linguistic interaction high acceptability depressive moods collection and analysis sational AI,
ing depressive (Malaysia, Philippines, related to depressive moods traits of depressive con- and entertainment, and engagement online, highlighting its employing
moods across dif- India, Indonesia, using specific keywords versations. Each user’s capable of generating globally application in social natural
ferent cultural Thailand) and 56,586 data was anonymized, responses based sciences research language
contexts conversations in Western focusing on non- on a vast database processing
countries (Canada, clinically diagnosed of user interactions to under-
United Kingdom, United users discussing depres- stand
States) sive emotions and respond
to user
queries
5 Dadi Population building proxy 11,175 participants considered participants who The combination The UKBB database The UKBB is openly Population modeling This complementarity of proxy machine No state
2021 [15] modeling measures by apply- have responded to cognitive of brain imaging is to date the most accessible with machine learning measures and original measures learning
ing machine learn- tests and questionnaires and pro- and target-specific extensive large-scale and has extensive data can derive measures at capturing multiple health-related
ing on multimodal vide ac-cess to their primary sociodemographic co-hort aimed at study- acquired on 500,000 of mental health constructs when modeling from,
MR images and rich demographic characteristics inputs often improved ing the determinants individuals aged from heterogeneous both, brain signals and sociodemo-
sociodemographic and brain im-ages approximation perfor- of the health outcomes 40–70 years covering inputs including brain graphic data
information mance in the general adult rich phenotypes, signals and question-
from the largest population health-related informa- naire data. This may
biomedical cohort tion, brain-imaging, complement or even
to date: the and genetic data substitute for psycho-
UK Biobank metric assessments
in clinical populations
Page 9 of 20
Table 3 (continued)
N Author Study type Study aims participants methodology analysis Meaning of Artificial Acceptability of application of Effectiveness of Artificial Kind of Name of Artificial
and year Intelligence Artificial Intelligence Artificial Intelligence Intelligence Artificial Intelligence
Intelligence

7 Rathnayaka designing Presenting 318 individuals The conceptual framework The conceptual frame- The conceptual frame- Users must vol-untarily Some of these The design and development of our chatbots Bunji
2022 [16, and developing the design across the world was formulated using a struc- work was designed work was designed submit their feelings responses were: BA-based AI chatbot, followed by its
30] of BA-based and development tured approach consistingof as a series of chatbot as a series of chatbot checks and PHQ2 1. “By using the app, participatory evaluation, confirmed
AI chatbot of BA-based AI three phases capabilities that can capabilities surveys while using I am more aware its effectiveness in providing support
and a pilot study chatbot, followed The conceptual framework be grouped into three the app of how my moods for individuals with mental health
by its participatory was designed as a series main categories, (1) per- fluctuate. It also made issues
evaluation in a pilot of chatbot capabilities that can sonalised conversation me think about what
Dehbozorgi et al. BMC Psychiatry

study setting be grouped into three main (2) emotional support I am grateful for,
that confirmed categories, (1) personalised con- and (3) remote mental which alleviated some
its effectiveness versation (2) emotional support health monitoring, negativity I was expe-
in providing support and (3) remote mental health which are followed riencing at the time”
for individuals monitoring, which are delineated by a pilot study Sensors 2022, 22, 3653
with mental health in the following subsections The 15 of 18
issues user experience of engaging 2. “Mood tracker
with the BA-based AI chatbot is great to be able
was a further design considera- to give to my medical
(2025) 25:132

tion that is articulated at the end practitioner—If my

of this section The participatory psychologist or doctor
evaluation of the completed asked me to use this
chatbot as a pilot study was con- as a monitoring tool
ducted in Australia in between sessions, I
would be more likely
to engage”
3. “Bunji lets me feel
courageous” them-
selves, and extending
the gamification
of Bunji for visually set-
ting and tracking goals
against a community
benchmark and one’s
own track record.
We are also working
on expanding
the activity banks
to include activities,
inspirations, quotations,
and workshops, as well
as enhancements
to the language model
for more fun/inspira-
tional human-like con-
versations with a range
of responses that suit
diverse demograph-
ics (e.g., emojis
and memes)
Page 10 of 20
Table 3 (continued)
N Author Study type Study aims participants methodology analysis Meaning of Artificial Acceptability of application of Effectiveness of Artificial Kind of Name of Artificial
and year Intelligence Artificial Intelligence Artificial Intelligence Intelligence Artificial Intelligence
Intelligence

6 Pei Two phase evaluating, predict- College students After determining the basic Data mining technol- The survey results By year, The computer technol- results show that there is no signicant Neural Not mentioned
2022 [17] ( Prediction ing, and analyzing algorithm used in the pre-diction ogy based on a neural show that from 2010 but the admission rate ogy of neural network correlation between college Students’ Network
and Analysis) the mental health and analysis of contemporary network algorithm to 2020,thenumber- has not changed sig- algorithm is applied inferiority complex and dependency
statusof contem- college students’ mental health, is used to collect data ofgraduatecandidate- nificantly Every year to the prediction map and the incidence of mental dis-
porary college was continued to test the accu- sources. Finally, the pre- shasincreasedyear of contemporary eases and majors. A comprehensive
students based racy of the algorithm, and finally, diction results are college students’ physical symptom test was con-
on a neural network determine an algorithm model analyzed, and the main mental health. A psy- ducted on individuals to understand
Dehbozorgi et al. BMC Psychiatry

algorithm with the highest prediction psychological stressor chological evaluation students’ psychological characteristics
accuracy factors of contemporary can evaluate students’ and behavior
college students are differences in learning
analyzed by cluster or ability, personality
analysis characteristics, relative
strengths, and weak-
nesses, evaluate
the development stage
that students have
(2025) 25:132

reached, determine
their relative strengths
and Weaknesses,
and find the reasons
for behavior changes
7 Tate Model develop- I.) development 7,638 twins The outcome, mental health Using a large range register informa- The highest ranking This top performing The strengths of this study include machine Not mentioned
2020 [18, ment a model that can from the Child and Ado- problems, was determined of data from parent tion, which can be variables were model would not be the comprehensive analysis of a wide learning
31] predict Mental lescent in Sweden by the Strengths and Dif- reports and register expensive or dif- either parent-rated suitable for clinical variety of factors associated with ado-
health problems at age 15 ficulties Questionnaire. Model data from numerous ficult for researchers or could easily be use, however it lays lescent mental health. Further,
in mid-adolescence performance was deter-mined Swedish national to obtain, may not be reported by parents, important groundwork the use of parental reports indicates
II.) investigating by the area under the receiver registers, this study pre- necessary for a suc- for future models that these risk factors are identifiable
if machine learning operating characteristic curve dicted adolescent men- cessful psychiatric risk seeking to predict by non-clinicians, indicating a low
techniques will (AUC) tal health reasonably model general mental health cost future solution for large scale
outperform logistic well, with a maxi-mum outcomes Future mental health screens
regression AUC of 0.739 on the test studies should make
set (using the random use of parent-rated
forest model). Although assessments when pos-
the AUC indi-cates sible. Additionally, it
an adequate model, it may not be necessary
is not accurate enough for similar studies
for clinical use to forgo logistic
regression in favor
of other more complex
methods
Page 11 of 20
Table 3 (continued)
N Author Study type Study aims participants methodology analysis Meaning of Artificial Acceptability of application of Effectiveness of Artificial Kind of Name of Artificial
and year Intelligence Artificial Intelligence Artificial Intelligence Intelligence Artificial Intelligence
Intelligence

8 Didarul empirical study to examine The target population Methodologically, non-parametric one Use of ANN to model Not explicitly discussed, ANN was used to com- ANN provided robust predictions Artificial Not specified
Alam et. al with the factors affecting of this study wasmHealth SEMANN sample Kolmogorov– and predict non- but implied to be plement PLS-SEM with small RMSE values, indicating Neural Net- by a particular brand
(2021) [19] SEMANN the intention users of any kind (Structural Equation Modeling Smirnov (K-S) test linear relationships positive as it was used in understanding linear high reliability and accuracy work (ANN) or model, referred
(Structural Equa- and actual in Bangladesh who were – Artificial Neural Network) Analysis of Variance in mHealth usage to enhance the study’s and non-linear associa- in predicting behavioral intentions with feed-for- to generally as ANN
tion Modeling usage behavior in self-quarantine dur- approach which can deal (ANOVA) test was con- and its determinants predictive accuracy tions among variables and mHealth usage ward-back-
– Artificial Neural on mHealth adop- ing COVID-19 pandemic with non-compensatory ducted propagation
Network tion, investigate and non-linear relationships And The sensitivity (FFBP)
Dehbozorgi et al. BMC Psychiatry

the effect of actual grounded on the UTAUT2 model analysis was done
usage behavior as a theoretical basis Pre-testing
of mHealth on men- was conducted with a pool
tal well-being of three industry ex perts, two
of the endusers, academicians, and five actual
and investigate users of mHealth services
the moderating role to confirm content validity. A few
of self-quarantine adjustments were made based
on the inten- on their suggestions. In addition,
(2025) 25:132

tion–actual a pilot testing comprising of 45

usage of mHealth actual mHealth users was admin-
under the coronavi- istered to check and ensure
rus disease (COVID- the reliability of the study
19) pandemic constructs
situation
9 Götzl et al. parallel mixed– To explore the atti- Young people (aged Qualitative: Focus groups Independent AI is understood Participants, AI is applied to person- The effectiveness of AI is seen in its The study Not specified
(2022) [20, method study tudes, preferences, 12–25), school psycholo- with young people and expert analysis of qualitative as systems that evalu- both young people alize intervention com- potential to provide personalized, focuses
28] and needs of young gists, psychological coun- interviews and quantitative data ate large amounts and experts, showed ponents in mHealth timely interventions and sup- on machine
people and key selors, media experts, Quantitative: Representative followed by integration of information to make a generally positive atti- apps, aiming to adapt port, although there are concerns learning
stakeholders and mHealth app online survey. to identify points decisions and learn tude towards the use to users’ specific needs about the technical background and algo-
towards AI- developers of convergence over time, often of AI in mHealth apps, and provide tailored and data privacy rithms
informed mHealth or divergence used in applications with a willingness recommendations as the types
apps and guide like YouTube, Spotify, to use AI-informed of AI used
the participatory and Google training if it is useful in mHealth
development and helps achieve apps
of such an app personal goals
10 Onuki. M Casual-compar- Estimating Physi- 97 participants who estimation, the heart rate Quantitative analysis wearable devices to Acceptable by users to notify one’s health Fitbit is estimated as maintaining Smart watch Fitbit Inspire HRTM
Et al ative research cal/Mental Health were all women and variability (HRV) compared through mathemati- measure heat rate for measuring blood- conditions, and a good quality of data for a long heat rate (a kind of smart
2022 [21] Condition Using lived in Japan. They with measuring the HRV, cal estimation variability (HRV) e.g., related data on smart for that purpose, period of time. and predict the variability watch)
Heart Rate Data were members of Peer predict the subjects’ physical/ smart watches, smart watches, e.g., the the level of stress subjects’ physical/mental health (HRV) ana-
From a Wearable Ring, which is a social mental health only from the shirts, chest strap Apple Watch or the based on heat rate according to lyzer
Device media site especially HR measured by Fitbit instead monitors, and so on Fitbit, due to its ease variability (HRV) used
for women’s cancer of the HRV of integration and for estimating these
survivors. All of the usability conditions
participants had
experienced some
kind of female-specific
cancers, but they are
as energeticas healthy
people now
11 Alamgir Multi-methods to enhance an AI 10 the hybrid strategy called the four major phases A system of Auto- It is acceptable for to identify the The efficiency of the model proved A tool for Facial expression
FM et al study to identify the Japanese female Deep Belief Rain Optimization including pre-process- matic recognition of users in mental facial expressions through the simulations and it automatic recognition system
2022 facial expressions participants (DBRO) technique ing, feature extrac- facial expressions health care of the individuals identified to outperform the other classifica-
of the individuals tion, feature selection and categorizing it existing approaches tion of
and categorizing and classification into seven different the facial
it into different emotions emotions
emotions
Page 12 of 20
Table 3 (continued)
N Author Study type Study aims participants methodology analysis Meaning of Artificial Acceptability of application of Effectiveness of Artificial Kind of Name of Artificial
and year Intelligence Artificial Intelligence Artificial Intelligence Intelligence Artificial Intelligence
Intelligence

12 Yanqi Guo, Experimen- To analyze the effect A total of 320 Japanese The control group received tradi- Statistical analysis AI technology in family The application of AI AI technology is used - improved the psychological Mobile agent Agent Technology
2023 tal study of psychological middle school students, tional family health education (SPSS 22.0 to perform education refers to sys- in family education for personalized health of the experimental group technology family education
(Randomized health based on arti- randomly divided - The experimental group descriptive statistics, tems that simulate is considered worthy support in family improve the level of middle school - Intelligent system
controlled trials) ficial intelligence into an experimental received mental health education independent sample human intelligence, of research and offers education systems. students’ mental health, which can learning
agent technology group and a control based on the Agent Technology t-tests and ANOVA) providing personalized advantages in auton- AI Products can help improve student forced symptoms, systems
on the imple- group, with 160 cases family education system showing significant learning experiences, omy, responsiveness, students carry out per- interpersonal tension and sensitivity, covering
mentation effect in each group - Basic information and mental psychological health improving students’ initiative, and mobility sonalized learning. depression, anxiety, learning pressure, all aspects
Dehbozorgi et al. BMC Psychiatry

of Japanese family health scores of both groups improvements in the AI- self-learning abilities, The unique teaching maladjustment, emotional imbalance, of personal-
education.—To were compared based group and broadening their mode greatly improves psychological imbalance, and many ized learning,
build personalized knowledge students’ concentration other psychological states includ-
support for the fam- and provides more -Personalized AI support outper- ing interac-
ily education system convenient and intel- formed traditional family education tive modules
based on mobile ligent Channels methods based on mobile agents and knowl-
agents for students to acquire has the advantages of autonomy, edge bases
knowledge and infor- responsiveness, initiative, and mobil-
mation, Thus improving ity, which provides a new idea
(2025) 25:132

students’ self-learning for family education

ability and broadening
their knowledge
- It includes educa-
tional apps and soft-
ware, online learning
platforms, virtual reality
(VR), augmented reality
(AR), intelligent tutor-
ing systems, and data
analytics tools
Page 13 of 20
Table 3 (continued)
N Author Study type Study aims participants methodology analysis Meaning of Artificial Acceptability of application of Effectiveness of Artificial Kind of Name of Artificial
and year Intelligence Artificial Intelligence Artificial Intelligence Intelligence Artificial Intelligence
Intelligence

13 Ilona Halim, To test Total of 36 young adult - Participants completed 2 iVR - Within-group analyses The study focuses Usability: Participants personalize the iVR -A significant increase in self-compas- AI Not specified
Lehan whether a novel participants sessions spaced 2 weeks apart with paired-samples on Virtual Reality (VR) rated the AI-integrated experience, sion was observed after a single iVR for enhanc-
Dehbozorgi et al. BMC Psychiatry

Stemmet, type of individual- - Participants were - Validated measures of self- 2-tailed t tests rather than Artificial iVR system highly, adapting scenarios session, demonstrating the potential ing
Sylvia Hach, ized virtual reality recruited from a univer- compassion and depression - Changes in Self- Intelligence (AI) with a SUS score and interactions based effectiveness of AI-driven iVR interactions
Richard (iVR) can improve sity community social were completed at baseline Compassion Scale (SCS) -Virtual Reality (VR) of 75.9, indicating good on the user’s responses in enhancing self-compassion with avatars
Porter, Hai- self-compassion media site and upon completion of each and Patient Health is an emerging technol- to excellent usability and progress -A strong correlation in virtual
Ning Liang, and decrease - They were aware iVR session Questionnaire (PHQ-8) ogy that can simulate User Experience: to enhance engage- between increased self-compassion environ-
Atiyeh depressive that the study - Additional measures were scores were analyzed realistic and immersive UEQ ratings showed ment and therapeutic and reduced depressive symptoms ments
Vaezipour, symptoms was investigating a treat- administered to assess partici- - Pearson correlations experiences strengths in novelty outcomes was found throughout the study, and possibly
Julie D - To evaluate ment for depression pants’ perceptions about usability were calculated to test within a virtual world and stimulation, Interactive Avatars: aligning with existing literature incorporat-
Henry, Nilu- the usability but were not recruited and acceptability of the iVR how participants’ outperforming 75% Using AI to develop that self-compassion can protect ing more
(2025) 25:132

far Baghaei, and acceptability based on a depression approach upon completion self-compassion of benchmark systems. more sophisticated, against depression strengthen
2023 [24, of this iVR approach diagnosis of both sessions and depressive symp- However, dependability interactive avatars partici-
33] as rated by partici- toms related to one was rated below aver- that can respond pant–virtual
pants another age, suggesting users to user inputs environment
Self-compassion experienced issues and emotions, creating interactions
and depressive symp- with control and sys- a more immersive
toms were assessed tem performance and supportive
using validated Participant Feedback: environment
measures at baseline Qualitative feedback Behavior Analysis:
and at the end of each highlighted the sys- AI can be utilized
iVR session tem’s ease of use, to analyze user behav-
- Within-group immersive experience, ior and emotional
analyses showed and positive user responses during ses-
significant increases interface design. sions, providing
in self-compassion However, participants real-time feedback
and non-significant desired more robust and adjustments
trends for reduced scenarios, improved to the therapeutic
depressive symptoms interaction with avatars, process
- Quantitative and quali- and enhanced Predictive Analytics:
tative feedback performance Leveraging AI
supported the iVR to predict user needs
approach as being and potential mental
acceptable and usable health issues, allowing
for proactive adjust-
ments and interven-
tions within the iVR
system
Enhanced Data
Collection: AI can
enhance the collection
and analysis of data
from iVR sessions, offer-
ing deeper insights
into user progress
and the effectiveness
of the intervention,
which can inform
future developments
and personalized treat-
ment plans
Page 14 of 20
Table 3 (continued)
N Author Study type Study aims participants methodology analysis Meaning of Artificial Acceptability of application of Effectiveness of Artificial Kind of Name of Artificial
and year Intelligence Artificial Intelligence Artificial Intelligence Intelligence Artificial Intelligence
Dehbozorgi et al. BMC Psychiatry

Intelligence

14 Naveen –––- an efficient deep ––––––– Experiments are conducted algorithm to under- Facial emotion facial emotion recogni- Facial emotionrec- the proposed facial emo-tion facial deep learning-based
Kumari· convolutional neural by using the benchmark stand and model recognition extracts tion model performs ognitionrepresents recognition model achieves better emotion facial emotion recog-
Rekha network model dataset and competitive human the relationships the human emotions considerably better the content of an input performance than the existing emo- recognition nition model
Bhatia is proposed to rec- emotion recognition models between faces from the images compared to the com- image in the form tion recognition models. Comparative model
2022 [25] ognize the human The contrast-limited adaptive and facial expressions and videos petitive models of human emotions analysis reveals that the proposed
emotions from facial histogram equaliza- tion (CLAHE) and to recognize by using various facial emotion recognition model
(2025) 25:132

images is applied to improve the visibility human emotions machines and deep achieves better results than the com-
of input images. The modified learning models petitive models in terms of accuracy
joint trilateral filter is applied
to the obtained enhanced
images to remove the impact
of noise
Finally, the deep convolutional
neural network was applied
on the CLAHE and modified joint
trilateral filter (MJTF)- based fea-
ture matrix. Next, Adam optimizer
was used to optimize the cost
function of deep convolutional
neural networks
15 Rohit The article Measuring 20 people experiment gathered Artificial see a drastic decre- make computersmimi- The use of Artificialintelligence AI and ML
Rastogi, is a trial enough Happiness Index around 50to60 different intelligence(AI)is ment in the value chumanbehaviorto (AI)in the field of health care can Expert
et al., to establish and Electronic radiation levels of six the ability of a com- of the radiation level makedecisions.Suc bedefined by System
2022 [26, the effect of Sans Gadgets Radiations categories of gadgets puterprogram of different electronic hthingsinvolveProb- the use of difficult andvarious algo- Big Data
34] kritVedicMantraa on AI IoT Systems: using various sensors or a machine to think gadgets lemSolvingwithout rithms along with software that are Healthcare
ndYajna,anancie Return to Indian and instruments and learn. traditionallycodingt- used for stimulating human cognition 4.0
ntHindu science Scriptures and Sci- and plotted a curve heactualsolutionofthep for analyzing and observing compli- IoT
of up liftment ence for Mental among the mas we roblem.Itcanalsobedefi cated medical data. In simpler words Machine
of consciousness Fitness During can see the trends nedtoanalyzethegiven it isthe ability of computer algorithms Learning
and to heal self Global Threats that the radiation dataandpredictbase- to approximate conclusions without- Artificial Intel-
by sound science level is greater dontheobservationvari- direct human input. Themaindiffer- ligence
and heatand before the Yagya ousaspectsaboutthedat ence between Traditional technolo-
light than the radiation level a.Recentlythesetechniq gies and AI technologist abilitytocolle
after performing Yagya ues areusedinthefieldo ctinformationordata,processitandgive
on the compound. fselfdrivingcars,humans awell- structuredresulttotheend user.
peechrecognition,data AI usemachine learning algorithm
mining,voiceassistance, stoper form the abovementioned
andmanymore.Allthes- tasks and operations. Recognition
etechnologiesaremade of patterns in behavior and creation
moreeasiertoimplemen of theirown logiccanbedone by i
tthroughthetermAI mplementingthesealgorithms.AI
algorithmsneedtobetestedrepeatedly-
forreducingthemargin of error.
Page 15 of 20
Dehbozorgi et al. BMC Psychiatry (2025) 25:132 Page 16 of 20

In summary, the overall quality of studies indicates a Didarul Alam et al. (2021) focused on mHealth users in
promising trend in AI applications within mental health Bangladesh during the COVID-19 pandemic, emphasiz-
research, with studies rated as “good” by recognized ing how global crises affect mental health and the role of
assessment tools, although some studies show moderate technology in delivering support [19].
methodological rigor, suggesting areas for improvement Götzl et al. (2022) included young people aged 12–25,
in future research designs. alongside stakeholders in mHealth app development, to
explore attitudes towards AI-informed mental health
Demographics of participants applications, indicating a collaborative approach to men-
In a comprehensive analysis of participant demographics tal health solutions [28].
across various studies examining the application of AI in Onuki et al. (2022) focused on 97 women cancer sur-
mental health, several key insights emerge. Inkster et al. vivors in Japan, providing insights into the mental health
(2018) focused on anonymous global users of the Wysa challenges faced by this specific demographic and the
app, specifically targeting individuals who self-reported potential for AI to offer tailored support [21]. Alamgir
symptoms of depression. This study provided insights et al. (2022) involved 10 Japanese female participants in
into how AI-enabled conversational tools can engage facial expression recognition studies, emphasizing the
users in mental health support [10]. intersection of AI and emotional recognition in mental
Xiao Li (2023) examined elderly participants from health contexts [32].
two community groups, a control group (n = 46) and an Yanqi Guo (2023) studied 320 Japanese middle school
intervention group (n = 47), to understand the factors students, comparing the effects of AI-based family edu-
influencing depression among older adults and to formu- cation systems on mental health, which highlights the
late psychological intervention plans. This demographic importance of early intervention in educational settings
highlights the importance of addressing mental health in [23].
aging populations [12]. Halim et al. (2023) recruited 36 young adult partici-
Liang et al. (2022) involved 185 college students in pants from a university community to evaluate the effects
sports majors, divided into training and test sets, to eval- of individualized virtual reality (VR) on self-compassion
uate the effectiveness of AI in assessing and improving and depression, showcasing innovative approaches to
mental health. The focus on college students underscores mental health interventions [33].
the relevance of mental health interventions in educa- Kumari and Bhatia (2022) did not specify participant
tional settings [13]. demographics but focused on deep learning for emotion
Chin et al. (2023) analyzed user interactions with the recognition, indicating the technical aspects of AI appli-
SimSimi chatbot, utilizing a dataset from 96,197 conver- cations in mental health [25].
sations in Eastern countries and 56,586 conversations in Finally, Rastogi et al. (2022) involved 20 participants
Western countries. This cross-cultural analysis provides in a trial related to ancient practices for mental fitness,
valuable insights into how different populations express reflecting a unique blend of traditional and modern
depressive moods through AI platforms [29]. approaches to mental health [34]. Overall, this diverse
Dadi (2021) utilized a large cohort from the UK range of participant demographics across studies high-
Biobank, comprising 11,175 participants, to study cogni- lights the multifaceted nature of mental health research
tive tests and demographic characteristics, emphasizing and the potential of AI to address the needs of various
the potential of AI in population modeling and mental populations. The findings underscore the importance of
health research [15]. tailoring AI applications to specific demographic groups
Rathnayaka (2022) conducted a pilot study with 318 to enhance their effectiveness and acceptability in mental
individuals globally to evaluate a BA-based AI chatbot health interventions.
designed for mental health support, demonstrating the
growing interest in AI applications across diverse popu- Acceptancy of AI in mental health
lations [30]. Pei (2022) focused on college students, In reviewing the acceptability of AI in various men-
although specific demographic details were not provided, tal health and healthcare contexts, studies demonstrate
emphasizing the need for targeted mental health inter- a general positive trend across different demograph-
ventions in this age group [17]. ics, applications, and intervention types. Inkster et al.
Tate (2020) studied 7,638 twins from a Swedish ado- (2018) show that the Wysa app, aimed at depression sup-
lescent mental health study, providing a unique perspec- port, demonstrated high engagement and effectiveness,
tive on genetic and environmental influences on mental emphasizing the potential acceptability of AI among
health outcomes [31]. users with self-reported depressive symptoms [10]
(Table 2).
Dehbozorgi et al. BMC Psychiatry (2025) 25:132 Page 17 of 20

In this view AI tools are especially beneficial in student signals from wearable devices to detect stress and anxi-
populations, where they effectively address mental health ety, enhancing real-time monitoring and intervention
challenges, particularly in sports education contexts. capabilities [17].
Moreover, advanced AI mechanisms, like attention- In healthcare systems, AI facilitates patient engage-
based LSTM models, have shown promise in improving ment and clinical decision-making. Guo et al. (2024) used
the accuracy of emotion recognition, an essential compo- AI to analyze wearable data for stress detection, while
nent in mental health interventions [27]. Götzl et al. (2022) incorporated AI into clinical docu-
In higher education, the feasibility of AI-driven mental mentation and patient interaction systems to improve
health plans for college students has been confirmed with clinician efficiency and patient satisfaction. Further-
favorable results, underscoring its practicality in aca- more, the application of AI in personalized mHealth apps
demic environments [13]. allows adaptation to user needs and preferences, enhanc-
The global scale of AI adoption is notable, with plat- ing intervention outcomes [23, 28].
forms reaching millions of users across 111 languages, Collectively, these studies highlight the diverse func-
underscoring a high level of acceptability and engage- tions of AI in mental health, encompassing real-time
ment [29]. Similarly, openly accessible databases like the monitoring, preventive care, personalized interventions,
UKBB, with extensive health data, signify the potential and predictive modeling. A common thread throughout
for AI integration in broader health-related applications these applications is AI’s capacity to bridge gaps in men-
[15]. Rathnayaka (2022) emphasizes that user-driven data tal health services, enhance patient engagement, and pro-
input, such as voluntary mental health surveys, aligns vide customized support for various populations.
well with AI’s non-invasive monitoring applications [30].
In youth-focused applications, Tate (2020) notes that
AI-driven interventions in mental health must address Effectiveness of AI in mental health
ethical concerns such as inequality and population bias The reviewed studies demonstrate the varied effective-
[31]. Similarly, Guo et al. (2024) suggest that wearable ness of AI in improving mental health assessment, moni-
devices provide an accepted, practical approach to moni- toring, and treatment, leading to notable advancements
toring stress and anxiety levels in real-time [23]. across different mental health conditions (Table 2).
These findings collectively highlight the high accept- In the domain of digital mental health interventions
ability and potential efficacy of AI-based mental health (DMHIs), Inkster et al. (2018) reported that high engage-
interventions across various settings, populations, and ment with the Wysa app led to significant improvements
applications. in self-reported depression scores among users, with a
moderate effect size [10].
Application of AI in mental health AI’s capacity to improve early mental health recog-
The studies highlight the extensive applications of AI nition and intervention is also evident in studies by Lei
across various domains in mental health and healthcare, et al. (2023) and Xiao Li (2023), where deep learning and
with each study exploring unique facets of AI’s function- attention mechanisms were used to significantly enhance
ality and its impact on patient outcomes (Table 2). emotion recognition and classification accuracy in men-
The effectiveness of AI in educational settings is fur- tal health monitoring [12, 27]. Furthermore, studies by
ther highlighted by Lei et al. (2023), where AI was used Liang (2022) emphasize the effectiveness of AI in psychi-
to reduce anxiety and depression symptoms among stu- atric assessment, including real-time symptom tracking
dents. In another innovative application, Xiao Li (2023) and personalized therapy delivery [13].
optimized LSTM outputs through AI attention mecha- At the same time, Tornero-Costa et al. (2023) empha-
nisms, which enhanced emotion recognition accuracy, sized the potential of AI in facilitating large-scale mental
a significant step for personalized mental health care health screenings and interventions, which could allow
[12, 27]. for the early identification of mental health issues and
Beyond individual assessments, AI is also used in pop- provide tailored support for various populations [1].
ulation-level mental health monitoring, as demonstrated In summary, these studies confirm the multifaceted
in study by Liang (2022) who used AI for psychiatric effectiveness of AI in mental health across preventive,
assessments, adherence tracking, and CBT delivery [13]. diagnostic, and therapeutic stages, with applications that
AI applications extend beyond therapy and assess- include emotion recognition and personalized mental
ments; they are increasingly relevant in preventive health interventions. This broad applicability under-
interventions and public health efforts. AI’s potential scores AI’s potential to enhance access to mental health
for predictive modeling is also recognized, with Pei resources, improve treatment delivery, and facilitate
(2022) applying AI to analyze physiological and digital
Dehbozorgi et al. BMC Psychiatry (2025) 25:132 Page 18 of 20

large-scale mental health screening and monitoring health challenges and the potential for AI to address
initiatives. these varied needs. Research targeting college students,
elderly populations, and individuals from diverse cultural
Discussion backgrounds demonstrates the adaptability of AI tools
The integration of AI into mental health care has in meeting the specific requirements of different groups.
emerged as a focal point of research, reflecting a wide This adaptability is essential for enhancing the acceptabil-
array of applications and methodologies. A systematic ity and effectiveness of AI interventions, as evidenced by
review encompassing 15 studies revealed varying degrees positive user feedback in studies like that of Götzl et al.
of methodological rigor and quality, offering a nuanced (2022), which explored young people’s attitudes toward
perspective on AI’s role in enhancing mental health AI-informed mHealth applications [28].
interventions. Notably, studies that employed robust Nevertheless, despite the promising applications of AI
methodologies and high-quality assessment tools, such in mental health, several ethical considerations and meth-
as the MMAT and the JBI checklist, provided more cred- odological limitations warrant attention. The reviewed
ible insights into the effectiveness of AI applications in studies highlighted concerns regarding data privacy,
this field. algorithm transparency, and the necessity for stakeholder
This review captures the evolution and current land- involvement in the development of AI technologies. As
scape of AI applications in mental health care, particu- Tavory (2023) noted, establishing ethical frameworks is
larly focusing on advancements from 2009 to 2024. crucial for guiding the responsible deployment of AI in
For instance, the research conducted by Didarul Alam mental health care. AI is increasingly becoming an inte-
et al. (2021) utilized a rigorous empirical framework, gral component of digital medicine, with the potential to
achieving a high-quality score of 10. This study ana- significantly influence mental health research and prac-
lyzed mobile health (mHealth) technologies during the tice. To fully leverage AI’s capabilities, it is essential for a
COVID-19 pandemic and demonstrated a positive cor- diverse community of stakeholders—including scientists,
relation between mHealth usage and mental well-being, clinicians, regulators, and patients—to engage in open
highlighting the potential of AI-driven tools to enhance communication and collaboration [5].
user engagement during crises [19]. Similarly, studies by As AI techniques continue to evolve, there is poten-
Inkster et al. (2018) and Chin et al. (2023), both of which tial to redefine mental illnesses more objectively than
received quality ratings of 7, showcased the effectiveness the current DSM-5 classification system. Such advance-
of AI chatbots like Wysa in improving mental health out- ments could enable earlier identification of mental ill-
comes through increased user engagement [10, 29]. nesses, even at prodromal stages, when interventions are
Conversely, studies with moderate quality scores, such likely to be more effective. Moreover, AI can facilitate the
as those by Dadi (2021) and Rathnayaka (2022), while customization of prescribed treatments based on indi-
still informative, indicated areas needing methodological vidual characteristics, leading to more personalized and
improvement. Dadi’s population modeling approach gar- effective mental health care. It is also noteworthy that
nered a score of 6, suggesting that although it provided consumers recognize several advantages in using medi-
valuable insights into the relationship between AI and cal chatbots, such as anonymity and quicker access to
mental health, its methodological limitations may hinder pertinent information. Previous research indicates that
the generalizability of its findings. Similarly, Rathnayaka’s consumers are often just as willing to share emotional
study on a behavioral activation-based AI chatbot also and personal information with a chatbot as they would
received a score of 6, underscoring the necessity for more with a human friend. Studies suggest that interactions
robust evaluation frameworks to assess the efficacy of with chatbots and humans yield comparable levels of per-
such interventions [15, 30]. ceived understanding, disclosure closeness, and cognitive
The findings from high-quality studies emphasize the reappraisal, indicating that individuals engage psycho-
critical importance of methodological rigor in AI-related logically with chatbots similarly to their interactions with
mental health research. For example, Xiao Li (2023) people. This reinforces the potential of chatbots to pro-
employed advanced AI techniques, including Long Short- vide effective support in mental health contexts. In this
Term Memory (LSTM) networks, to enhance emotion systematic review, we synthesized evidence regarding the
recognition, achieving a quality score of 7 [12]. This illus- effectiveness and user evaluation of AI-based conversa-
trates that higher-quality studies not only yield more reli- tional agents (CAs) in mental health care. Our findings
able data but also contribute to the advancement of AI suggest that these CAs can effectively alleviate psycho-
methodologies in mental health applications. logical distress, with the most significant effects observed
Furthermore, the diversity of participant demographics in studies utilizing generative AI, multimodal or voice-
across the studies reflects the complex nature of mental based CAs, and interventions delivered through mobile
Dehbozorgi et al. BMC Psychiatry (2025) 25:132 Page 19 of 20

applications and instant messaging platforms. This aligns Supplementary Information

with the findings of Shimada (2023) and Alhuwaydi The online version contains supplementary material available at https://​doi.​
(2024), who highlight AI’s potential in addressing dispari- org/​10.​1186/​s12888-​025-​06483-2.
ties in access to mental health services [35, 36].
Supplementary Material 1.
These findings are further supported by Oladimeji et al.
(2023), who emphasize AI’s role in the early detection
and prevention of mental health issues [37]. However, it Acknowledgements
Special thanks go to the vice-chancellor of the Community Based Psychiatric
is crucial to recognize that these technologies could exac- Care Research Center, School of Nursing and Midwifery, Shiraz University of
erbate existing inequalities if not implemented thought- Medical Sciences.
fully and equitably. Addressing potential disparities in
Human ethics and consent to participate declarations
access and ensuring that all populations benefit from AI The study did not involve direct human participants, and therefore, specific
advancements in mental health care is essential to pre- declarations regarding human ethics and consent to participate are not
vent widening the gap in mental health services. applicable.

Clinical trial number

Conclusion This study did not involve clinical trials, and as such, there is no clinical trial
number to report.
The paper provides a comprehensive review of the entire
spectrum of AI in mental health, highlighting its posi- Financial disclosure
tive contributions to the field. AI holds numerous prom- No.
ises for enhancing mental health care, and this paper Author’s contributions
explores various facets of its application. AI technologies RD, NJM, EKH, SZ, and, FL were involved in Conceptualization, and MY and MV
are anticipated to introduce innovations to existing medi- collected the dates. RD, FHH, DHN and PS analyzed the data. RD, SZ, EKH, DHN,
HRH, PS, MY, FHH, MV, NJM, and FL were involved in the methodology. NJM,
cal practices and future health care systems. Currently and RD was project administrator and supervised the project. RD and NJM
available AI-based health care technologies have demon- validated all stages of the project. RD, MV, MY, EKH, SZ, PS, FHH and, FL wrote
strated significant efficacy in accurately diagnosing and the original draft of the manuscript and review and editing was done by RD,
NJM, EKH, SZ,HRH.
classifying patient conditions, as well as predicting dis-
ease trajectories by leveraging accumulated medical data. Funding
No.

Limitation Data availability

One significant limitation of this systematic review is On reasonable request, the corresponding author is willing to provide the
datasets used and analyzed during the present study.
the variability in methodological quality and reporting
standards among the included studies. Although vari-
Declarations
ous quality assessment tools were employed, such as
the Mixed Methods Appraisal Tool (MMAT) and the Ethics approval and consent to participate
Newcastle–Ottawa Scale (NOS), discrepancies in study This research was conducted in accordance with the ethical standards set
forth in the Declaration of Helsinki. The findings will be presented in a collec-
design, sample sizes, and data collection methods were tive format to ensure transparency and accessibility.
observed. Many studies lacked transparency in report-
ing AI model features and data preprocessing tech- Competing interests
The authors declare no competing interests.
niques, which are crucial for ensuring reproducibility
and reliability. This inconsistency may lead to biases in Author details
1
the interpretation of AI effectiveness in mental health Community Based Psychiatric Care Research Center, School of Nursing
and Midwifery, Shiraz University of Medical Sciences, Shiraz, Iran. 2 Namazi
applications and limit the generalizability of the find- Hospital, Shiraz University of Medical Sciences, Shiraz, Iran. 3 Health Deputy,
ings across diverse populations. Kermanshah University of Medical Science, Kermanshah, Iran. 4 Department
Future research should prioritize the establishment of Nursing, Shiraz Branch, Islamic Azad University, Shiraz, Iran. 5 Department
of Psychiatric Nursing, School of Nursing & Midwifery, Isfahan University
of standardized methodologies and reporting guide- of Medical Sciences, Isfahan, Iran. 6 University of Science and Technology
lines for AI applications in mental health. This would (IUST), Tehran, Iran. 7 Department of Educational Sciences, Faculty of Education
enhance the robustness of the evidence base, facilitate and Psychology, Azarbaijan Shahid Madani University, Tabriz, Iran. 8 Master’s
Student in Clinical Psychology, Faculty of Medical Sciences, Hamedan Branch,
comparative analyses, and ultimately improve the inte- Islamic Azad University, Hamedan, Iran. 9 Nursing and Midwifery Care Research
gration of AI technologies into clinical practice. Addi- Center, Isfahan University of Medical Sciences, Isfahan, Iran. 10 Department
tionally, involving a broader range of stakeholders, of Psychology, Faculty of Psychology, Payam Noor University, Kaboudar Ahang
Center, Hamedan, Iran. 11 School of Nursing and Midwifery, Shiraz University
including mental health professionals and patients, in of Medical Sciences, Shiraz, Iran.
the development and evaluation of AI tools may help
address ethical concerns and ensure that these technol- Received: 21 November 2024 Accepted: 7 January 2025
ogies meet the needs of diverse user groups.
Dehbozorgi et al. BMC Psychiatry (2025) 25:132 Page 20 of 20

References 22. Alamgir FM, Alam MS. An artificial intelligence driven facial emotion
1. Image ; David Novillo-Ortiz2 RT-CAOIAM-MAOINA-MAOILLAOIVTAO. recognition system using hybrid deep belief rain optimization. Multimed
Methodological and Quality Flaws in the Use of Artificial Intelligence in Tools Appl. 2023;82(2):2437–64.
Mental Health Research: Systematic Review. JMIR mental health. 2023;10. 23. Guo Y. Examining the impact of assistive technology on psychological
https://​doi.​org/​10.​2196/​42045 health, family education, and curriculum research in Japan: insights from
2. Nilsen P, Svedberg P, Nygren J, Frideros M, Johansson J, Schuel- artificial intelligence. J Autism Dev Disord. 2024;54(10):3823–36.
ler S. Accelerating the impact of artificial intelligence in mental 24. Halim I, Stemmet L, Hach S, Porter R, Liang H-N, Vaezipour A, et al. Indi-
healthcare through implementation science. Implement Res Pract. vidualized virtual reality for increasing self-compassion: evaluation study.
2022;3:26334895221112030. JMIR Mental Health. 2023;10:e47617.
3. Gaffney H, Mansell W, Tai S. Conversational agents in the treatment of 25. Kumari N, Bhatia R. Efficient facial emotion recognition model using
mental health problems: mixed-method systematic review. JMIR Mental deep convolutional neural network and modified joint trilateral filter. Soft
Health. 2019;6(10):e14166. Comput. 2022;26(16):7817–30.
4. Olawade DB, Wada OZ, Odetayo A, David-Olawade AC, et al. Enhancing 26. Rastogi R, Chaturvedi D, Saxena M, Sagar S, Gupta M, Choudhary R, et al.
mental health with Artificial Intelligence: Current trends and future pros- Measuring happiness index and electronic gadgets radiations on AI IoT
pects. Journal of Medicine, Surgery, and Public Health. 2024;3. https://​doi.​ systems: return to Indian scriptures and science for mental fitness during
org/​10.​1016/j.​glmedi.​2024.​100099 global threats. Int J Soc. 2022;13:37.
5. Tamar T. Regulating AI in mental health: ethics of care perspective. JMIR 27. Lei L, Li J, Li W. Assessing the role of artificial intelligence inthe mental
Mental Health. 2024;11:e58493. https://​doi.​org/​10.​2196/​58493. healthcare of teachers and students. Soft Computing. 2023:1–11.
6. Tornero-Costa R, Martinez-Millana A, Azzopardi-Muscat N, Lazeri L, Traver 28. Götzl C, Hiller S, Rauschenberg C, Schick A, Fechtelpeter J, Fischer Abaigar
V, Novillo-Ortiz D. Methodological and quality flaws in the use of artificial U, et al. Artificial intelligence-informed mobile mental health apps for
intelligence in mental health research: systematic review. JMIR Mental young people: a mixed-methods approach on users’ and stakeholders’
Health. 2023;10:e42045. https://​doi.​org/​10.​2196/​42045. perspectives. Child Adolesc Psychiatry Ment Health. 2022;16(1):86.
7. Hong QN, Fàbregues S, Bartlett G, Boardman F, Cargo M, Dagenais P, et al. 29. Chin H, Song H, Baek G, Shin M, Jung C, Cha M, et al. The potential of
The Mixed Methods Appraisal Tool (MMAT) version 2018 for information chatbots for emotional support and promoting mental well-being
professionals and researchers. Educ Inf. 2018;34(4):285–91. in different cultures: mixed methods study. J Med Internet Res.
8. Wells G. The Newcastle-Ottawa Scale (NOS) for assessing the quality of 2023;25:e51712.
nonrandomised studies in meta-analyses department of epidemiol- 30. Rathnayaka P, Mills N, Burnett D, De Silva D, Alahakoon D, Gray R. A
ogy and commuunity medicine, University of Ottawa,Room 3227A, 451 mental health chatbot with cognitive skills for personalised behavioural
Smyth Road, Ottawa, Ontario K1J 8M5. Canada: NOS; 2024. [updated activation and remote health monitoring. Sensors. 2022;22(10):3653.
2021; cited 2024]. [email protected]. Available from: https://​www.​ohri.​ 31. Tate AE, McCabe RC, Larsson H, Lundström S, Lichtenstein P, Kuja-Halkola
ca/​progr​ams/​clini​cal_​epide​miolo​gy/​oxford.​asp. R. Predicting mental health problems in adolescence using machine
9. Jordan Z, Lockwood C, Munn Z, Aromataris E. The updated Joanna learning techniques. PLoS ONE. 2020;15(4):e0230389.
Briggs Institute model of evidence-based healthcare. JBI Evid Implement. 32. Alamgir FM, Alam MS. An artificial intelligence driven facial emotion rec-
2019;17(1):58–71. ognition system using hybrid deep belief rain optimization. Multimedia
10. Inkster B, Sarda S, Subramanian V. An empathy-driven, conversational Tools Appl. 2023;82(2):2437–64.
artificial intelligence agent (Wysa) for digital mental well-being: real- 33. Halim I, Stemmet L, Hach S, Porter R, Liang H-N, Vaezipour A, et al. Indi-
world data evaluation mixed-methods study. JMIR Mhealth Uhealth. vidualized virtual reality for increasing self-compassion: evaluation study.
2018;6(11):e12106. JMIR Mental Health. 2023;10:e47617.
11. Lei L, Li J, Li W. Assessing the role of artificial intelligence in the mental 34. Rastogi R, Chaturvedi D, Saxena M, Sagar S, Gupta M, Choudhary R, et al.
healthcare of teachers and students. Soft Computing. 2023;2023:1–11. Measuring happiness index and electronic gadgets radiations on AI IoT
12. Li X. Evaluation and analysis of elderly mental health based on artificial systems: return to Indian scriptures and science for mental fitness during
intelligence. Occup Ther Int. 2023;2023(1):7077568. global threats. Int J Soc Ecol Sustain Dev (IJSESD). 2022;13(1):1–37.
13. Liang L, Zheng Y, Ge Q, Zhang F. Exploration and strategy analysis of men- 35. Shimada K. The role of artificial intelligence in mental health: a review. Sci
tal health education for students in sports majors in the era of artificial Insights. 2023;43(5):1119–27.
intelligence. Front Psychol. 2022;12:762725. 36. Alhuwaydi AM. Exploring the role of artificial intelligence in mental
14. Chin H, Song H, Baek G, Shin M, Jung C, Cha M, et al. The potential of healthcare: current trends and future directions–a narrative review for a
chatbots for emotional support and promoting mental well-being in comprehensive insight. Risk Manage Healthc Pol. 2024;17:1339–48.
different cultures: mixed methods study. J Med Int Res. 2023;25:e51712. 37. Oladimeji KE, Nyatela A, Gumede S, Dwarka D, Lalla-Edward ST. Impact of
15. Dadi K, Varoquaux G, Houenou J, Bzdok D, Thirion B, Engemann D. Popu- artificial intelligence (AI) on psychological and mental health promotion:
lation modeling with machine learning can enhance measures of mental an opinion piece. New Voices Psychol. 2023;13:12 pages- pages.
health. GigaScience. 2021;10(10):giab071.
16. Rathnayaka P, Mills N, Burnett D, De Silva D, Alahakoon D, Gray R. A
mental health chatbot with cognitive skills for personalised behavioural Publisher’s Note
activation and remote health monitoring. Sensors. 2022;22(10):3653. Springer Nature remains neutral with regard to jurisdictional claims in pub-
17. Pei J. Prediction and analysis of contemporary college students’ mental health lished maps and institutional affiliations.
based on neural network. Comput Intell Neurosci. 2022;2022(1):7284197.
18. Tate AE, McCabe RC, Larsson H, Lundström S, Lichtenstein P, Kuja-Halkola
R. Predicting mental health problems in adolescence using machine
learning techniques. PloS one. 2020;15(4):e0230389.
19. Alam MMD, Alam MZ, Rahman SA, Taghizadeh SK. Factors influencing
mHealth adoption and its impact on mental well-being during COVID-19
pandemic: A SEM-ANN approach. J Biomed Inform. 2021;116:103722.
20. Götzl C, Hiller S, Rauschenberg C, Schick A, Fechtelpeter J, Fischer Abaigar
U, et al. Artificial intelligence-informed mobile mental health apps for
young people: a mixed-methods approach on users’ and stakeholders’
perspectives. Child Adolesc Psychiatry. 2022;16:86.
21. Onuki M, Sato M, Sese J, editors. Estimating physical/mental health
condition using heart rate data from a wearable device. 2022 44th
Annual International Conference of the IEEE Engineering in Medicine &
Biology Society (EMBC); 2022 11-15 July 2022; Glasgow, Scotland. United
Kingdom: IEEE; 2022.