344P Towards robust computation of AI/ML computational pathology IHC scoring with the radial digital tumour proportion score (rdTPS)

Abstract

Background: Global cancer incidence is expected to grow from 19.3 m in 2020 to 28.4 m by 2040, emphasizing the need for scalable long-term care.In breast cancer, delayed follow-up for early-stage cases and poor adherence monitoring in advanced stages worsen outcomes.We developed China's leading AI-powered follow-up platform for breast cancer, integrating the ChatGLM model with national guidelines, patient education, and real-world data.The system drives a physician mini-program using virtual personas and cloned voices to automate expert-level patient interaction and follow-up care. Methods:We conducted 2 investigator-initiated clinical studies: 1) early-stage breast cancer-measuring follow-up adherence and quality of life; 2) advanced-stage patients on ADC therapy-evaluating treatment adherence and adverse event monitoring.Platform functions included emotional support, breast cancer-specific Q&A, risk assessment, reminders, report interpretation, rehab planning, follow-up protocols, and record summaries.It currently operates in 60 top-tier hospitals across 24 provinces, with full coverage in Gansu, Ningxia, and Yunnan.6 expert interviews were conducted to evaluate implementation.Results: AI-based medication guidance reduced initial counseling time by 60%, and report automation cut consultation volume by 50%.The digital rehab module increased patient education reach by 45%.Proactive adverse event monitoring improved satisfaction by 35%.Smart reminders lowered rates of missed doses and follow-ups.Real-time symptom capture and automated report processing boosted follow-up efficiency and coordination. Conclusions:This platform improves patient satisfaction, reduces unscheduled visits, and enhances operational efficiency.It supports hierarchical care and expert sharing across hospital levels, while enabling decentralized research via remote participation and broader geographic and population inclusion.Our model represents a scalable and transferable framework for AI-assisted oncology care in diverse healthcare systems globally.

Ähnliche Arbeiten