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

Abstract

Abstract The study introduces a clinical decision support system (CDSS) developed at a single academic cancer center, integrating real-time clinical, genomic, and imaging data for over 170,000 patients across 11 cancer types. We have developed the Yonsei Cancer Data Library (YCDL) data integration framework to continuously collect and update multimodal datasets comprising over 800 features per case. Quality control measures, using 143 logical comparisons, addressed missing data and outliers, achieving median accuracies of 92.6% for surgical and 98.7% for molecular pathology. An Extract-Transform-Load (ETL) process with natural language processing transformed unstructured data, enabling survival analyses stratified by tumor stage, which revealed significant stage-dependent differences. The CDSS dashboard visualizes patient trajectories and key milestones. User feedback from oncology professionals showed strong acceptance, with satisfaction scores exceeding 4 out of 5. This framework demonstrates the potential of multimodal data integration to enhance clinical decision-making and patient outcomes, with future research needed to validate its generalizability and scalability.