A comprehensive survey on medical concept normalization: Datasets, techniques, applications, and future directions.

Area of Science:

• Medical Informatics
• Natural Language Processing
• Computational Linguistics

Background:

• Medical Concept Normalization (MCN) maps informal medical terms to standardized concepts, vital for medical text analysis and healthcare intelligence.
• Existing research has focused on datasets and algorithms but lacked a holistic survey of MCN.
• This gap necessitates a comprehensive review of MCN's core components and applications.

Purpose of the Study:

• To provide the first comprehensive survey of Medical Concept Normalization (MCN).
• To cover MCN task definitions, annotation schemes, datasets, normalization techniques, performance, and applications.
• To offer guidelines for model selection and discuss future research directions.

Main Methods:

• Systematic review and comparative analysis of existing MCN literature and datasets.
• Quantitative evaluation of various deep learning models (CNN, RNN, Transformer, GNN, LLM) on benchmark datasets.
• Exploration of MCN applications in EHR management, clinical decision support, precision medicine, and health information exchange.

Main Results:

• A detailed overview of MCN task definitions, annotation schemes, and available datasets.
• Comparative analysis of benchmark datasets and leading normalization methods.
• Quantitative performance evaluation of diverse MCN models, highlighting state-of-the-art approaches.

Conclusions:

• This survey offers a valuable resource for researchers and practitioners by consolidating MCN knowledge.
• It identifies key challenges and outlines promising future research directions in the field.
• The provided GitHub repository offers access to related resources for further exploration.