Estimating personalized risk ranking using laboratory test and medical knowledge (UMLS)

Area of Science:

• Medical Informatics
• Computational Biology
• Clinical Decision Support

Background:

• Physicians face challenges in interpreting complex laboratory test results.
• Accurate and timely disease diagnosis is crucial for effective patient care.
• Existing diagnostic tools may lack personalization based on individual patient data.

Purpose of the Study:

• To introduce the Concept Graph Engine (CG-Engine) for personalized disease ranking.
• To leverage laboratory test data and a medical knowledge base for improved diagnostic insights.
• To enhance physician efficiency by providing a clear overview of patient health status.

Main Methods:

• Developed a Concept Graph Engine (CG-Engine) utilizing the Unified Medical Language System (UMLS).
• Constructed a two-level concept graph linking laboratory tests to diseases.
• Computed weighted paths between tests and diseases using attributes like relation types and semantic types.
• Aggregated path weights to generate personalized disease rankings.

Main Results:

• The CG-Engine successfully generates patient-specific disease rankings based on laboratory data.
• The system effectively utilizes the UMLS as a comprehensive medical knowledge base.
• The engine computes weights between laboratory tests and diseases, enabling nuanced analysis.
• Personalized disease rankings are derived by aggregating weighted paths.

Conclusions:

• The CG-Engine offers a novel approach to personalized disease ranking from laboratory data.
• This tool can significantly improve physician throughput by providing concise diagnostic snapshots.
• The system enhances clinical decision-making through personalized and data-driven insights.