Continuous multimodal data supply chain and expandable clinical decision support for oncology

Area of Science:

• Oncology
• Medical Informatics
• Data Science

Background:

• Clinical decision-making in oncology is complex, often requiring integration of diverse patient data.
• Existing systems may not effectively leverage multimodal data, including clinical, genomic, and imaging information.
• Academic cancer centers face challenges in managing and analyzing large-scale patient datasets.

Purpose of the Study:

• To develop and evaluate a clinical decision support system (CDSS) integrating multimodal data for cancer patient care.
• To establish a robust data integration framework (Yonsei Cancer Data Library - YCDL) for continuous data collection and updating.
• To assess the impact of multimodal data integration on clinical decision-making and patient outcome analysis.

Main Methods:

• Developed the Yonsei Cancer Data Library (YCDL) data integration framework for multimodal datasets (clinical, genomic, imaging).
• Implemented quality control measures (143 logical comparisons) for data accuracy (surgical: 92.6%, molecular pathology: 98.7%).
• Utilized an Extract-Transform-Load (ETL) process with natural language processing for data transformation and survival analyses stratified by tumor stage.

Main Results:

• The YCDL framework successfully integrated multimodal data for over 170,000 patients across 11 cancer types.
• Survival analyses revealed significant stage-dependent differences in patient outcomes.
• The CDSS dashboard effectively visualizes patient trajectories and key milestones.
• Oncology professionals reported high satisfaction (scores >4/5) with the CDSS.

Conclusions:

• Multimodal data integration via the YCDL framework significantly enhances clinical decision-making in oncology.
• The developed CDSS demonstrates potential for improving patient outcomes through data-driven insights.
• Future research should focus on validating the generalizability and scalability of this integrated data framework.