Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Ya-Mei Bai has a total of 204 co-authors. Here are the top co-authors by number of shared publications:

  • Mu-Hong Chen

    Department of Psychiatry, Taipei Veterans General Hospital, Taiwan

    50 shared publications
  • Tung-Ping Su

    39 shared publications
  • Shih-Jen Tsai

    Department of Psychiatry, Taipei Veterans General Hospital, Taipei, Taiwan; Department of Psychiatry, College of Medicine, National Yang Ming Chiao Tung University, Taipei, Taiwan

    36 shared publications
  • Ju-Wei Hsu

    Department of Psychiatry, Taipei Veterans General Hospital, Taipei, Taiwan; Department of Psychiatry, College of Medicine, National Yang Ming Chiao Tung University, Taipei, Taiwan

    32 shared publications
  • Cheng-Ta Li

    31 shared publications
  • Mu-Hong Chen

    Department of Psychiatry, Taipei Veterans General Hospital, Taipei, Taiwan; Division of Psychiatry, Faculty of Medicine, National Yang Ming Chiao Tung University, Taipei, Taiwan

    30 shared publications

Latest Publications

  • Open Access

    A Multimodal Learning Approach Based on Self-narrative for Cognitive Impairment Prediction in Patients with Unipolar and Bipolar Disorder: Methodology Development and Application (Preprint)

    by Ya-Han Hu, Cheng-Che Shen, Ya-Mei Bai, Ruei-Yan Wu

    BACKGROUND Mood disorders, including unipolar and bipolar disorder, are commonly associated with significant cognitive impairment (CI) that adversely affects patient recovery and functional outcomes. Although CI prediction research has advanced considerably in neurodegenerative disorders, robust predictive models integrating demographic variables with multimodal speech-based features remain lacking for mood disorder populations. OBJECTIVE This study aimed to develop and evaluate a multimodal framework combining demographic variables, audio features, and text features extracted from Chinese self-narrative to predict executive function performance as measured by the Wisconsin Card Sorting Test (WCST). We also sought to identify which feature types (handcrafted vs. learned) and modalities (audio, text, or combined) best predict specific cognitive domains. METHODS A dataset comprising demographic data, speech recordings, and WCST scores from 133 mood disorder patients (unipolar and bipolar disorder) and 63 healthy controls recruited from a hospital in northern Taiwan was analyzed. Audio features included handcrafted features (prosodic and extended Geneva Minimalistic Acoustic Parameter Set [eGeMAPSv01a] features) and learned embeddings (VGGish, Wav2Vec 2.0, HuBERT), while text features included handcrafted features (term frequency–inverse document frequency [TF-IDF] and linguistic inquiry and word count [LIWC]) and learned embeddings (TAIDE, ERNIE-Health, RoBERTa). Five classifiers were evaluated across three experimental configurations (audio + demographic, text + demographic, and multimodal), targeting four WCST measures: total errors (TE), perseverative responses (PR), perseverative errors (PE), and categories completed (CC). Model performance was assessed using the area under the receiver operating characteristic curve (AUC) as the primary metric, with paired t-tests used for comparative evaluation. RESULTS Statistical analysis identified psychiatric hospitalization history as significantly associated with all four WCST measures, while years of education and education level, age, marital status, alcohol consumption, hypertension, and hyperlipidemia showed domain-specific associations. Predictive modeling using extreme gradient boosting as the reference classifier revealed modality-specific advantages: audio-based models (prosodic features + demographic variables; AUC = 0.715) demonstrated superior performance for TE; text-based models (ERNIE-Health + demographic variables) excelled for PE (AUC = 0.727) and PR (AUC = 0.703); and multimodal models (prosodic + LIWC + demographic variables) achieved the best performance for CC (AUC = 0.759), significantly outperforming both unimodal and demographic-only baselines. CONCLUSIONS A multimodal framework integrating demographic, audio, and text features from brief clinical speech recordings demonstrates feasibility and modality-specific advantages for predicting CI in mood disorders. These findings support the development of objective, scalable cognitive screening tools applicable in clinical settings, with potential to facilitate rapid assessment and accessible self-monitoring for mood disorder patients. CLINICALTRIAL The current study received approval from the Institutional Review Board (IRB) of Taipei Veterans General Hospital (IRB number: SE22066A).

  • Risk of Neurodevelopmental Disorders Following the Herpesviral Infection in Preschooler Children

    by Li‐Chi Chen, Wei‐Sheng Huang, Ju‐Wei Hsu, Ya‐Mei Bai, Shih‐Jen Tsai, Tung‐Ping Su, Tzeng‐Ji Chen, Mu‐Hong Chen

    The associations between the herpesviral infections and neurodevelopmental disorders, including autism and attention deficit hyperactivity disorder (ADHD), remained unclear. Using the Taiwan National Health Insurance Research Database, the present study included 27,922 children < 6 years who were exposed to herpesviral infections between 1997 and 2011, as well as 111,688 age‐ and sex‐matched children who were not exposed. All participants who developed autism and ADHD, as well as started to use ADHD medications during the follow‐up (from enrollment to the end of 2013) were identified. Human herpesvirus (HHV) subtypes were further assessed for the neurodevelopmental disorder risk. After adjusting for demographic characteristics, perinatal conditions, Charlson Comorbidity Index scores, and all‐cause clinical visits, we found that any herpesvirus infection was associated with increased risks of developing autism (hazard risk [HR]: 2.23), ADHD (1.53), and ADHD medication use (HR: 1.42). We further observed such increased risks in HHV‐1/2, HHV‐3, HHV‐4, and other HHVs, but not in HHV‐5. Finally, a herpesviral infection involving the central nervous system was markedly related to higher risks of developing autism (HR: 6.16) and ADHD medication use (1.77). Preschool children exposed to herpesvirus infection had higher subsequent risks of autism and ADHD, as well as a greater likelihood of ADHD medication use than unexposed peers. Elucidating the immunoneuro developmental mechanisms underlying herpesviral infection and neurodevelopmental disorders will require additional investigation.

  • Domain- and item-level symptom thresholds for severe cognitive impairment in major depressive disorder

    by Kuan-I Liu, Wei-Chen Lin, Ju-Wei Hsu, Li-Chi Chen, Shih-Jen Tsai, Ya-Mei Bai, Mu-Hong Chen

  • Open Access

    Risks and Familial Coaggregation of Autism Spectrum Disorder Among First-Degree Relatives of Individuals With Colorectal Cancer

    by Dian-Jeng Li, Shih-Jen Tsai, Chih-Ming Cheng, Wen-Han Chang, Ya-Mei Bai, Tung-Ping Su, Tzeng-Ji Chen, Mu-Hong Chen, Chih-Sung Liang

    Objective Associations between major mental disorders (MMDs) and colorectal cancer (CRC) have been reported; however, it remains unknown whether this association extends to an individual’s kinship. The aim of this study was to explore the risks of MMDs in first-degree relatives (FDRs) of individuals with CRC.Methods We used data from the Taiwan National Health Insurance Research Database. FDRs of individuals with CRC were identified as the index group, and a demographic-matched group was also included as the controls. The primary outcome was the risk of an MMD. A Poisson regression model with robust error variance was used to estimate the relative risks and 95% confidence intervals.Results A total of 68,965 FDRs of individuals with CRC (38,024 males and 30,941 females) and 275,860 matched controls were included. After adjustments for potential confounders, the FDR-CRC group had a higher risk of autism spectrum disorder than the controls, especially in the males of fathers with CRC. In addition, individuals who had a son with CRC had a higher risk of bipolar disorder than the controls, especially females. Males whose mother had CRC had higher risks of attention deficit hyperactivity disorder and general anxiety disorder than the controls.Conclusion With the unique association identified in this study between FDRs of individuals with CRC and the risk of MMDs, genetic studies are encouraged to explore the complicated etiologies behind this association.

  • Functional dysconnectivity of large-scale functional brain networks in young adults with bipolar disorder with and without low-grade inflammation

    by Tien-Wei Hsu, Jia-Ru Li, Shih-Jen Tsai, Ya-Mei Bai, Tung-Ping Su, Tzeng-Ji Chen, Ju-Wei Hsu, Mu-Hong Chen, Chih-Sung Liang

  • Open Access

    Risk of major psychiatric disorder after adolescent pregnancy: A nationwide cohort study of 149,870 girls

    by Tien‐Wei Hsu, Jia‐Ru Li, Shih‐Jen Tsai, Ya‐Mei Bai, Tung‐Ping Su, Tzeng‐Ji Chen, Ju‐Wei Hsu, Mu‐Hong Chen, Chih‐Sung Liang

    Aim: Adolescent pregnancy is a global public health concern associated with increased risks of physical health, mental well‐being, and social functioning; however, its potential to increase the long‐term risk of major psychiatric disorder (MPD) remains underexplored. Methods: Between 1996 and 2013, a cohort of 29,974 girls with adolescent pregnancy without MPD before and 1:4 matched controls based on age, family income, residence location, and individual psychiatric comorbidity were enrolled. A Cox regression model was used, adjusted for potential confounders, to estimate the risk of schizophrenia, bipolar disorder, and major depressive disorder (MDD). Results: After adjusting for confounding factors, adolescent pregnancy was associated with significantly increased risks of bipolar disorder (hazard ratio 1.44; 95% confidence interval 1.20–1.72) and MDD (1.88; 1.76–2.00), but not schizophrenia (0.94; 0.73–1.21), compared to the control group. These associations persisted after excluding the first 3 years of follow‐up, with elevated risks of bipolar disorder (1.27; 1.05–1.54) and MDD (1.61; 1.50–1.72) among adolescents with pregnancy compared to controls. Conclusions: These results underscore the importance of implementing long‐term mental health surveillance strategies for young mothers who experience adolescent pregnancy.

  • Parent-of-origin effects and cross-disorder transmission in familial psychiatric risk: A nationwide study of 1.77 million offspring

    by Ta-Chuan Yeh, Chih-Wei Hsu, Shih-Jen Tsai, Ya-Mei Bai, Tung-Ping Su, Tzeng-Ji Chen, Chih-Sung Liang, Mu-Hong Chen, Tien-Wei Hsu


  • Open Access

    Role of glymphatic system function in cognitive dysfunction among patients with bipolar disorder

    by Mu-Hong Chen, Ying-Jay Liou, Ju-Wei Hsu, Shih-Jen Tsai, Chiung-Chih Chang, Ya-Mei Bai

    Objective: Increasing evidence has suggested a potential role of glymphatic system function in the pathomechanisms underlying mood disorders and related cognitive impairment. However, this association in bipolar disorder remained unclear. Design, Participants, and Measures: In all, 185 patients (mean age ~ 30 years) with bipolar disorder and 95 healthy individuals were enrolled for the assessment of the go/no-go task and the digit symbol substitution test (DSST). The Analysis Along the Perivascular Space (ALPS) index, an indicator of glymphatic system activity, was measured using diffusion tensor imaging within diffusion magnetic resonance imaging. All participants were divided into two subgroups (< vs. ≥ mean level) based on the mean ALPS index level. Additionally, participants were further categorized into quartiles based on ALPS index levels, ranging from Q1 (highest) to Q4 (lowest). Results: Generalized linear model, after adjusting for age, sex, education years, and clinical symptoms, found no difference in the ALPS index levels between patients and controls. Patients with lower ALPS index levels—that is, those with values below the mean or in the lowest quartile (Q4)—performed the worst on the DSST and the go/no-go task. Conclusions and Relevance: Glymphatic system dysfunction appeared to play a critical role in cognitive impairment among patients with bipolar disorder, but not in healthy individuals. Further research is warranted to clarify the specific pathomechanisms underlying this bipolar disorder–specific association between glymphatic system dysfunction and cognitive deficits.


  • Open Access

    Aspirin and atorvastatin augmentation treatment in patients with major affective disorders and inflammatory dysregulation: a 12-week randomised Placebo-controlled study

    by Ya-Mei Bai, Ju-Wei Hsu, Ying-Jay Liou, Mu-Hong Chen

    Background: This study investigated whether aspirin and atorvastatin provide additional antidepressant effects in patients with major affective disorders and inflammatory dysregulation. Methods: Three 12-week treatment groups, each receiving aspirin (100 mg/day), atorvastatin (10 mg/day), or a placebo, were randomly assigned to 14 patients (seven with major depressive disorder [MDD] and seven with bipolar disorder [BD]), as well as two additional groups of 17 patients (each with nine patients with MDD and eight patients with BD). All patients had Clinical Global Impressions scores ≤3 and met the criteria for inflammatory dysregulation (i.e., C-reactive protein (CRP) level ≥ 1,000 ng/ml or soluble tumour necrosis factor-α receptor 1 (TNF-αR1) level ≥ 800 pg/ml). The Hamilton Depression Rating Scale (HDRS) and Montgomery-Sberg Depression Rating Scale (MADRS) were used to assess depressive symptoms, and the Global Assessment of Functioning Scale (GAF) was used to assess overall functioning. Baseline and week 12 CRP, TNF-αR1, and soluble IL-2 receptor (sIL-2R) levels were evaluated. Results: Generalised estimating equation models demonstrated a reduction in total HDRS ( p < 0.001) and MADRS ( p < 0.001) scores and an increase in GAF scores ( p < 0.001) in the medication groups compared with the placebo group. Only atorvastatin increased anti-inflammatory cytokine sIL-2R levels ( p < 0.001). Both atorvastatin ( p < 0.001) and aspirin ( p = 0.025) raised proinflammatory cytokine sTNF-αR1 levels. Discussion: Aspirin and atorvastatin improved depressive symptoms and overall function in patients with major affective disorders. However, both medications raised TNF-αR1 levels, and only atorvastatin increased sIL-2R levels.