GUIDE-G: An artificial intelligence-powered platform for dynamic NCCN guideline visualization in breast cancer (BC).

Abstract

9027 Background: Breast cancer (BC) management has evolved rapidly, with 17 FDA approvals in the past five years. Increasing complexity in NCCN guidelines has resulted in prolonged decision-making times and complex clinic preparation processes. GUIDE-G, an AI-powered, web-based platform, was developed to address these challenges by enabling hierarchical visualization of NCCN guidelines and providing efficient, point-of-care clinical decision support. Methods: The primary objective was to assess the feasibility of GUIDE-G. SUMMARIZATION of NCCN Guidelines in BC was performed using Large Language Model-based feature extraction, applying targeted prompts for accurate content retrieval and generating structured markup output. Manual curation by a panel of BC experts at Baylor College of Medicine ensured alignment with the guidelines, with accuracy measured by edits per generated content. VISUALIZATION involved transforming the markup into dynamic hierarchical diagrams using markmap-lib framework, supported by Node.js. GitHub-based version control ensured automated updates, with mobile-compatible HTML enabling cross-platform accessibility. Results: GUIDE-G (https://elkhanany.github.io/cancer_workflow/) achieved 90% accuracy in guideline summarization, as validated by expert review. The feasibility objective was met, with seamless integration into clinical workflows. Anecdotal feedback from 20 BC fellows highlighted the platform's ability to reduce per-patient preparation time by approximately 5 min. Although formal survey data is pending, preliminary observations suggest the platform enhances learner engagement and supports real-time clinical decision-making. Visualization performance demonstrated consistent sub-second rendering on major mobile platforms, ensuring accessibility at the point of care. Formal impact evaluation is underway, with IRB approval for studies assessing workflow efficiency and learner satisfaction. Conclusions: GUIDE-G offers a transformative approach to BC guideline implementation, combining scalable architecture and version-controlled deployment for sustained adaptation to evolving guidelines. Learner impact is being collected via 16-point survey instruments. Preliminary findings demonstrate potential for significant improvements in clinical workflow efficiency and educational outcomes, paving the way for implementation across oncology subspecialties and transforming evidence-based care delivery.