سامانه همایش های دانشگاه آزاد اسلامی واحد نجف آباد

صفحه اصلی / اولین همایش بین المللی تحول خدمات سلامت: تلفیق مراقبت، درمان، بازتوانی و امیدبخشی

Intelligent Psychotherapy Navigation Agent: A Hybrid Intelligence and Human-in-the-Loop Framework for Mental Health Triage in Telemedicine

نویسندگان :

Hossein Tork Ladani¹ Shirin Salmanizadeh² Mahdi Tork Ladani³

1- دانشگاه آزاد واحد نجف آباد 2- دانشگاه ازاد نجف اباد 3- دانشگاه شاهد

کلمات کلیدی :

Telemedicine،Hybrid Intelligence،Human-in-the-Loop،Large Language Models،Mental Health Triage،Clinical Decision Support،Digital Mental Health

چکیده :

Background: The global burden of mental health disorders continues to rise, while access to psychological care remains constrained by workforce shortages and geographic barriers [1]. Telemedicine platforms increasingly incorporate artificial intelligence to extend service capacity [2], yet autonomous use of large language models (LLMs) in clinical settings raises concerns regarding reliability and accountability [3,4]. This study presents the design of an Intelligent Psychotherapy Navigation Agent within the “Doctor Agent” telemedicine ecosystem, structured around three distinct functions: triage, navigation, and structured intervention. Method: The framework combines a Hybrid Intelligence paradigm with a Human-in-the-Loop (HITL) architecture [5] across three care tiers. A multi-domain psychiatric intake form is processed through NLP-based extraction, converting patient-reported symptoms into structured clinical variables. These variables, rather than the language model itself, feed a deterministic clinical rule engine that performs tier assignment, with safety indicators always overriding severity-based routing. Finding: Patients are dynamically routed into Tier 1 (protocol-informed, CBT-based digital intervention for low-risk symptoms), Tier 2 (AI-assisted monitoring under active clinician supervision for moderate cases), and Tier 3 (immediate specialist referral and crisis escalation). Scenario-based simulation evaluation against expert-defined reference decisions showed promising concordance with clinician judgment and encouraging performance in identifying crisis scenarios. Conclusion: Continuous human oversight and transparent, rule-based justification preserve clinical safety while supporting clinician accountability. The agent functions as a triage, navigation, and structured-intervention tool rather than an autonomous therapist. Prospective clinical trials are required to validate these preliminary findings. Keywords: Telemedicine, Hybrid Intelligence, Human-in-the-Loop, Large Language Models, Mental Health Triage, Clinical Decision Support, Digital Mental Health