진단변증 스키마 기반 AI 멀티모달 임상 표현을 활용한 표준화 환자 구현 시스템 개발 = A Study on the Development of an AI-Based Standardized Patient System Utilizing Multimodal Clinical Representations Based on a Diagnostic Pattern Identification Schema

다국어 초록 (Multilingual Abstract)

This study proposes a novel approach to enhancing the clinical
competencies of Korean medicine students by expanding
standardized patient (SP) education in a manner that reflects the
unique diagnostic framework of Korean medicine. AI-based SP
systems capable of supporting visual clinical information
interpretation and pattern-based reasoning are emerging as
essential tools for competency-based education, improving
learning efficiency and broadening the scope of instruction.
To achieve this goal, we developed CURA, an AI standardized
patient system that integrates large language models (LLMs),
retrieval-augmented generation (RAG), and multimodal
technologies within a diagnostic and pattern-identification
framework. CURA dynamically presents visual materials during
natural, interactive dialogues with learners, enabling the use of
multidimensional diagnostic information such as patient
photographs, thermometer images, chest X-ray, and tongue
images. This system allows learners to comprehensively interpret
and analyze visual and verbal cues in accordance with Korean
medicine diagnostic logic and to engage in a clinical reasoning
process that closely mirrors real-world practice.
The CURA system supports training in the differentiation of
respiratory disease patterns, including the wind-cold and
wind-heat patterns associated with the common cold and allergic
rhinitis, and provides an immersive environment that enhances
clinical reasoning ability. It also generates clinical scenarios,
facilitates interactions with standardized patients, and produces
real-time conversational reports containing key symptoms and
patient histories. Furthermore, an automated assessment module
based on CPX scoring criteria leverages conversational and
observational data to deliver immediate scoring and feedback.
This feature enables learners to promptly identify omitted
history-taking items or diagnostic errors for self-directed
learning, while allowing instructors to evaluate students’ clinical
performance using consistent standards and thereby minimize
assessment bias. To evaluate the content validity and educational
utility of the CURA system, an expert review was conducted.
The evaluation comprised 15 items across three domains—content
validity, educational effectiveness, and usability—assessed
using a 5-point Likert scale. The results indicated mean
scores of 4.57 for content validity, 4.64 for educational
effectiveness, and 4.63 for usability, with all domains scoring
above 4.57. Experts particularly emphasized the system’s
contribution to enhancing learning comprehension through
the provision of multimodal information (mean=4.7), fostering
the development of clinical reasoning skills (mean=4.85), and
supplementing clinical education (mean=4.70). These findings
support the clinical and educational validity of the CURA
system and suggest its potential applicability in real-world
educational settings.
In conclusion, this study presents a model that advances the
digital transformation of clinical education in Korean medicine,
supports competency-based training for Korean medicine
students, and offers meaningful implications for the future
direction of Korean medicine education.

목차 (Table of Contents)

• Ⅰ. 서론 1
• Ⅱ. 연구 방법 13
• Ⅲ. 결과 28
• Ⅳ. 고찰 60
• Ⅴ. 결론 66
• 참고문헌 67