Asthma Diagnosis AI Model Development
AI model for asthma diagnosis using methacholine challenge test data from Samsung Medical Center, achieving comparable performance with a single test.
Period: 2021.03 – 2023.10
Affiliation: Samsung Medical Center, Department of Allergy and Clinical Immunology
Summary: Developed an AI model for asthma diagnosis using clinical data from Samsung Medical Center. While conventional diagnosis requires three separate tests, the model achieved statistically comparable performance using only one test by leveraging previously unused variables discovered through data mining.
Key Responsibilities:
- Designed inclusion & exclusion criteria in collaboration with clinicians
- Extracted data from CDW and built analysis database
- Data cleaning, EDA, and predictive variable generation from multiple data sources (lab data, questionnaire data, sputum test data, bronchial challenge test results)
- Developed ML-based asthma diagnosis model; verified model reliability with clinicians via Shapley value analysis
Outcome: 2 conference presentations, 1 SCI publication (featured in the journal’s Editorial)