# Mapping the Landscape of Medical AI Research in Korea Using Topic Modeling

## TL;DR

This study maps the evolution of medical AI research in Korea from 2015 to 2024, showing a shift from algorithmic focus to patient-centered and system-integrated applications.

## Contribution

The paper introduces a chronological analysis of Korean medical AI research using topic modeling and keyword network analysis.

## Key findings

- Three research stages were identified: Introduction (2015–2018), Expansion (2019–2022), and Post-ChatGPT (2023–2024).
- Dominant themes include diagnostic imaging, healthcare system integration, and patient-centered disease modeling.
- Research shifted from algorithm-focused to system-level and patient-oriented applications.

## Abstract

Background/Objectives: This study analyzed ten years of domestic research on medical artificial intelligence (AI) from 2015 to 2024 using topic modeling and keyword network analysis. Chronological comparison showed that the research emphasis evolved through three stages—Introduction (2015–2018), Expansion (2019–2022), and Post-ChatGPT (2023–2024)—reflecting the growing incorporation of AI into clinical and service domains. Methods: We collected a curated set of 686 papers from the Korea Citation Index (KCI). After preprocessing—stopword removal, synonym unification, and lemmatization—7489 unique terms were extracted for the analysis. Results: Topic modeling identified three dominant themes: Diagnostic Imaging and Algorithm Validation, Healthcare Service and System Integration, and Patient-Centered Prediction and Disease Modeling. Keyword network analysis further revealed a structural shift from algorithm-oriented studies to system-level and patient-focused applications. Conclusions: These findings indicate that Korean medical AI research is maturing toward a more interpretable, integrated, and human-centered paradigm, underscoring the need for explainable AI (XAI), multidisciplinary collaboration, and governance frameworks for safe and ethical deployment.