Scholarlink

Psychiatric disorders (PDs), such as schizophrenia (SZ) and bipolar disorder (BP), significantly impact global populations, yet their diagnosis remains heavily reliant on subjective clinical methods. This paper presents the PD diagnosis system that integrates wearable electrocardiogram (ECG) monitoring, Time Series Convolutional Attention Network, dual visual and textual explanations with Explainable AI (XAI) and Large Visual Language Models (LVLMs), and user-centered interfaces to ensure model transparency, increase applicability in clinical settings. By investigating the relationship between PDs and heart rate variability (HRV), this work paves the way for more objective and accessible clinical assessments with wide-ranging applications in mental health diagnostics. We evaluate our system on the detection of PDs, demonstrating superior performance compared to recent literature models and providing interpretable explanations for model decisions. Additionally, we showcase the system’s applicability to a related use case, highlighting its scalability and potential for widespread adoption.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Human-Centered Explainable Psychiatric Disorder Diagnosis System Using Wearable ECG Monitors

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Hung Nguyen,
  • Alireza Rahimi,
  • Veronica Whitford,
  • Hélène Fournier,
  • Irina Kondratova,
  • René Richard,
  • Hung Cao

摘要

Psychiatric disorders (PDs), such as schizophrenia (SZ) and bipolar disorder (BP), significantly impact global populations, yet their diagnosis remains heavily reliant on subjective clinical methods. This paper presents the PD diagnosis system that integrates wearable electrocardiogram (ECG) monitoring, Time Series Convolutional Attention Network, dual visual and textual explanations with Explainable AI (XAI) and Large Visual Language Models (LVLMs), and user-centered interfaces to ensure model transparency, increase applicability in clinical settings. By investigating the relationship between PDs and heart rate variability (HRV), this work paves the way for more objective and accessible clinical assessments with wide-ranging applications in mental health diagnostics. We evaluate our system on the detection of PDs, demonstrating superior performance compared to recent literature models and providing interpretable explanations for model decisions. Additionally, we showcase the system’s applicability to a related use case, highlighting its scalability and potential for widespread adoption.