Jingyu Lei
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Jingyu Lei has a total of 47 co-authors. Here are the top co-authors by number of shared publications:
- Zhengzhi Feng
Department of Medical Psychology, Army Medical University
4 shared publications - Wei Li
Experimental Research Center for Medical and Psychological Sciences, School of Psychology, Third Military Medical University
3 shared publications - Xuerong Liu
Experimental Research Center of Medical and Psychological Science (ERC-MPS), School of Psychology, Army Medical University, Chongqing, China
3 shared publications - Zhiyi Chen
Third Military Medical University
3 shared publications Chang Shen
Experimental Research Center of Medical and Psychological Science (ERC-MPS), College of Psychology, Army Medical University, Chongqing, 400038, China
2 shared publications- Fangjian Liu2 shared publications
Latest Publications
- Open Access
- 15 September 2026
- Published by JMIR Publications Inc. in JMIR Formative Research
- Vol. 10, e92646
A Smartphone-Based Acoustic Machine Learning Pipeline for Detecting Suicidal Ideation: Case-Control Model Development and Validation Study
Background: Suicidal ideation (SI) among university students is a growing public health concern. Self-report screening can be limited by concealment and delayed disclosure. We evaluated a leakage-resistant, proof-of-concept pipeline to detect SI from standardized smartphone-recorded speech. Objective: This study aimed to extract acoustic markers from brief smartphone-based reading tasks and develop machine learning models for suicide risk prediction in university students, enabling low-cost, scalable early screening to support campus mental health services. Methods: Questionnaire data and speech recordings were collected via a WeChat mini program. After screening and clinical confirmation, 96 participants (n=48, 50% with SI; n=48, 50% controls) were included. Age and sex were evaluated as potential confounders. Each participant read 16 standardized sentences. Acoustic features were extracted using openSMILE (version 3.0.2), yielding a 570D feature vector per utterance. To prevent leakage from multiple recordings per speaker, we used participant-level 5-fold cross-validation, assigning all recordings from each participant to a single fold. Seven machine learning algorithms were evaluated using area under the curve (AUC), accuracy, and F 1 -score. Results: Acoustic-based models discriminated participants with SI from control participants. The SI group was significantly older than the control group ( P =.001). Random forest achieved an AUC of 0.813 (accuracy=0.748), and naive Bayes achieved an AUC of 0.806 (accuracy=0.757). Feature families related to pitch, mel-frequency cepstral coefficients, and harmonicity contributed to model performance. Conclusions: Standardized read speech captured via smartphones shows preliminary feasibility for SI discrimination under a leakage-aware evaluation design. External validation and testing with more naturalistic speech are warranted.
- 15 September 2026
- Published by JMIR Publications Inc. in JMIR Formative Research
- Vol. 10, e92646
Open AccessA Smartphone-Based Acoustic Machine Learning Pipeline for Detecting Suicidal Ideation: Case-Control Model Development and Validation Study
Background: Suicidal ideation (SI) among university students is a growing public health concern. Self-report screening can be limited by concealment and delayed disclosure. We evaluated a leakage-resistant, proof-of-concept pipeline to detect SI from standardized smartphone-recorded speech. Objective: This study aimed to extract acoustic markers from brief smartphone-based reading tasks and develop machine learning models for suicide risk prediction in university students, enabling low-cost, scalable early screening to support campus mental health services. Methods: Questionnaire data and speech recordings were collected via a WeChat mini program. After screening and clinical confirmation, 96 participants (n=48, 50% with SI; n=48, 50% controls) were included. Age and sex were evaluated as potential confounders. Each participant read 16 standardized sentences. Acoustic features were extracted using openSMILE (version 3.0.2), yielding a 570D feature vector per utterance. To prevent leakage from multiple recordings per speaker, we used participant-level 5-fold cross-validation, assigning all recordings from each participant to a single fold. Seven machine learning algorithms were evaluated using area under the curve (AUC), accuracy, and F 1 -score. Results: Acoustic-based models discriminated participants with SI from control participants. The SI group was significantly older than the control group ( P =.001). Random forest achieved an AUC of 0.813 (accuracy=0.748), and naive Bayes achieved an AUC of 0.806 (accuracy=0.757). Feature families related to pitch, mel-frequency cepstral coefficients, and harmonicity contributed to model performance. Conclusions: Standardized read speech captured via smartphones shows preliminary feasibility for SI discrimination under a leakage-aware evaluation design. External validation and testing with more naturalistic speech are warranted.
- Open Access
- 8 January 2026
- Published by JMIR Publications Inc. in JMIR Preprints
Robust Acoustic Markers of Suicidal Ideation in College Students: A Multiverse Analysis of Machine Learning Pipelines (Preprint)
BACKGROUND Suicide is a major global public health crisis and a leading cause of unnatural death among college students. Current suicide risk assessment mainly relies on self-report questionnaires and structured interviews. These methods are vulnerable to response bias and cannot support continuous monitoring. There is an urgent need for objective and non-invasive markers of suicide risk. Speech provides a promising source of such markers, as acoustic features reflect emotional and cognitive states related to suicidal ideation. Although speech-based machine learning models have shown encouraging predictive performance, most studies rely on single analytical pipelines. Consequently, the robustness and generalizability of reported acoustic markers across analytical choices remain a question for clinical translation. OBJECTIVE This study aimed to systematically evaluate how analytical variability influences predictive performance and feature stability in acoustic-based suicidal ideation detection. In addition, we sought to identify robust vocal biomarkers that generalize across analytical specifications. METHODS A comprehensive multiverse analysis was conducted across 1,764 distinct analytical pipelines using speech data from 96 Chinese university students (48 clinically confirmed cases of suicidal ideation and 48 matched controls). The pipelines varied in preprocessing strategies, acoustic feature sets, dimensionality reduction methods, and machine learning models. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC). Feature importance was aggregated across all pipelines to identify the top 10 core acoustic features. These features were subsequently examined within a new multiverse analysis framework to assess their robustness across analytical specifications. RESULTS Predictive performance was highly sensitive to analytical choices, with AUC values ranging from near chance (0.508) to high discriminative accuracy (0.856). Despite this variability, a core subset of acoustic features—including fundamental frequency (F0), F0 envelope, and Mel-frequency cepstral coefficients (MFCCs)—demonstrated robust and stable associations with suicidal ideation. These features remained statistically significant in 238 of 240 eligible specifications (99.2%). CONCLUSIONS Although speech-based computational prediction of suicidal ideation is highly dependent on analytical decisions, the underlying discriminative acoustic signals are remarkably stable. Multiverse analysis offers a transparent and rigorous framework for distinguishing robust vocal biomarkers from pipeline-dependent findings. This approach supports the development of reliable, interpretable, and clinically deployable voice-based mental health screening tools. CLINICALTRIAL This research was registered prospectively in the Chinese Clinical Trial Registry (http://www.chictr.org.cn) with the registration number ChiCTR2500106625.
- 8 January 2026
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessRobust Acoustic Markers of Suicidal Ideation in College Students: A Multiverse Analysis of Machine Learning Pipelines (Preprint)
BACKGROUND Suicide is a major global public health crisis and a leading cause of unnatural death among college students. Current suicide risk assessment mainly relies on self-report questionnaires and structured interviews. These methods are vulnerable to response bias and cannot support continuous monitoring. There is an urgent need for objective and non-invasive markers of suicide risk. Speech provides a promising source of such markers, as acoustic features reflect emotional and cognitive states related to suicidal ideation. Although speech-based machine learning models have shown encouraging predictive performance, most studies rely on single analytical pipelines. Consequently, the robustness and generalizability of reported acoustic markers across analytical choices remain a question for clinical translation. OBJECTIVE This study aimed to systematically evaluate how analytical variability influences predictive performance and feature stability in acoustic-based suicidal ideation detection. In addition, we sought to identify robust vocal biomarkers that generalize across analytical specifications. METHODS A comprehensive multiverse analysis was conducted across 1,764 distinct analytical pipelines using speech data from 96 Chinese university students (48 clinically confirmed cases of suicidal ideation and 48 matched controls). The pipelines varied in preprocessing strategies, acoustic feature sets, dimensionality reduction methods, and machine learning models. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC). Feature importance was aggregated across all pipelines to identify the top 10 core acoustic features. These features were subsequently examined within a new multiverse analysis framework to assess their robustness across analytical specifications. RESULTS Predictive performance was highly sensitive to analytical choices, with AUC values ranging from near chance (0.508) to high discriminative accuracy (0.856). Despite this variability, a core subset of acoustic features—including fundamental frequency (F0), F0 envelope, and Mel-frequency cepstral coefficients (MFCCs)—demonstrated robust and stable associations with suicidal ideation. These features remained statistically significant in 238 of 240 eligible specifications (99.2%). CONCLUSIONS Although speech-based computational prediction of suicidal ideation is highly dependent on analytical decisions, the underlying discriminative acoustic signals are remarkably stable. Multiverse analysis offers a transparent and rigorous framework for distinguishing robust vocal biomarkers from pipeline-dependent findings. This approach supports the development of reliable, interpretable, and clinically deployable voice-based mental health screening tools. CLINICALTRIAL This research was registered prospectively in the Chinese Clinical Trial Registry (http://www.chictr.org.cn) with the registration number ChiCTR2500106625.
- 20 November 2025
- Published by Elsevier in Journal of Affective Disorders
- Vol. 395 (Pt A) , 120636
Early identification of pediatric depression in western China: A multicenter, citywide evaluation of nine machine learning models
- 20 November 2025
- Published by Elsevier in Journal of Affective Disorders
- Vol. 395 (Pt A) , 120636
Early identification of pediatric depression in western China: A multicenter, citywide evaluation of nine machine learning models
- Open Access
- 28 August 2025
- Published by IMR Press in ALPHA PSYCHIATRY
- Vol. 26 (4)
Perceived Parental Involvement Decreases the Risk of Adolescent Depression
Objective: To tailor culturally sensitive interventional strategies for safeguarding adolescents’ mental health, this study investigated the role of perceived parental involvement in predicting depressive symptoms among Chinese adolescents, considering family socioeconomic status (SES). Methods: A cluster convenience sampling method recruited 21,818 participants from 48 middle schools across 29 provinces in China. The perceived parental involvement (PPI) Scale and the Chinese version of the center for epidemiologic studies depression scale (CES-D) assessed parental involvement and depressive symptoms, respectively. Data analysis employed linear mixed-effect models (LMM) and latent profile analysis (LPA). Results: The results indicated that 35.26% of adolescents exhibited subclinical depressive symptoms. LMM analysis revealed that higher perceived parental involvement scores, particularly emotional involvement, significantly predicted lower CES-D scores (β = –0.45, p < 0.001). LPA identified three distinct family factors profiles, with the “High SES-High PPI” group showing the lowest depression scores. Conclusion: The findings underscore the protective benefits of perceived parental involvement, especially emotional support, in mitigating depressive symptoms among adolescents. Specifically, adolescents from families with both high SES and high parental involvement exhibited the lowest levels of depressive symptoms, suggesting that interventions should focus on enhancing emotional support and addressing socioeconomic disparities to effectively reduce adolescent depression.
- 28 August 2025
- Published by IMR Press in ALPHA PSYCHIATRY
- Vol. 26 (4)
Open AccessPerceived Parental Involvement Decreases the Risk of Adolescent Depression
Objective: To tailor culturally sensitive interventional strategies for safeguarding adolescents’ mental health, this study investigated the role of perceived parental involvement in predicting depressive symptoms among Chinese adolescents, considering family socioeconomic status (SES). Methods: A cluster convenience sampling method recruited 21,818 participants from 48 middle schools across 29 provinces in China. The perceived parental involvement (PPI) Scale and the Chinese version of the center for epidemiologic studies depression scale (CES-D) assessed parental involvement and depressive symptoms, respectively. Data analysis employed linear mixed-effect models (LMM) and latent profile analysis (LPA). Results: The results indicated that 35.26% of adolescents exhibited subclinical depressive symptoms. LMM analysis revealed that higher perceived parental involvement scores, particularly emotional involvement, significantly predicted lower CES-D scores (β = –0.45, p < 0.001). LPA identified three distinct family factors profiles, with the “High SES-High PPI” group showing the lowest depression scores. Conclusion: The findings underscore the protective benefits of perceived parental involvement, especially emotional support, in mitigating depressive symptoms among adolescents. Specifically, adolescents from families with both high SES and high parental involvement exhibited the lowest levels of depressive symptoms, suggesting that interventions should focus on enhancing emotional support and addressing socioeconomic disparities to effectively reduce adolescent depression.
- Open Access
- 28 October 2024
- Published by Center for Open Science in OSF Preprints
Risk of bias and low reproducibility of meta-analytic estimates on mental health problems epidemics during the covid-19 pandemic: a meta-research analysis of global evidence on 18,604,876 individuals across 94 countries
First-tracking publications to covid-19-related meta-analytic evidences benefited evidence-based policymaking for safeguarding public mental health in this pandemic. However, such makeshifts had been concerned to cause troubling unreliable estimates for mental health problems epidemics. From 98 meta-analytic studies with 18,604,876 individuals across 94 countries published in the pandemic, we found significant risk of bias (ROBs) in publication, with one meta-analysis for being published per five days at the peak. Despite large-scale samples in these meta-analytic evidences, third-nation populations were substantially underrepresented. Of these meta-analyses, 70.6% (57.1%) showed high ROBs in transparency (methodology), and less than one-fifth of them could be empirically reproducible. Publication in high-impact journals, gaining high citations and fast-tracked publication did not improve transparency/reproducibility. By using individual participant data after minimizing these ROBs, we finally re-estimated these mental health problems epidemics to establish benchmarks. This study indicated that research quality/integrity should be still leading when compromising in emergency conditions.
- 28 October 2024
- Published by Center for Open Science in OSF Preprints
Open AccessRisk of bias and low reproducibility of meta-analytic estimates on mental health problems epidemics during the covid-19 pandemic: a meta-research analysis of global evidence on 18,604,876 individuals across 94 countries
First-tracking publications to covid-19-related meta-analytic evidences benefited evidence-based policymaking for safeguarding public mental health in this pandemic. However, such makeshifts had been concerned to cause troubling unreliable estimates for mental health problems epidemics. From 98 meta-analytic studies with 18,604,876 individuals across 94 countries published in the pandemic, we found significant risk of bias (ROBs) in publication, with one meta-analysis for being published per five days at the peak. Despite large-scale samples in these meta-analytic evidences, third-nation populations were substantially underrepresented. Of these meta-analyses, 70.6% (57.1%) showed high ROBs in transparency (methodology), and less than one-fifth of them could be empirically reproducible. Publication in high-impact journals, gaining high citations and fast-tracked publication did not improve transparency/reproducibility. By using individual participant data after minimizing these ROBs, we finally re-estimated these mental health problems epidemics to establish benchmarks. This study indicated that research quality/integrity should be still leading when compromising in emergency conditions.