时 间：2025-06-06 （周五）13:30 -14:15
报告人: 林存洁 中国人民大学副教授
地 点: 普陀校区理科大楼A1514
主持人: 章迎莹 华东师范大学副教授
摘 要：
Time-to-event data play a pivotal role in precision medicine, yet numerous existing studies neglect the presence of cured patients. When estimating the optimal Dynamic Treatment Regime (DTR) with a cure fraction, it is essential to strike a balance between the cure rate and the survival time of uncured patients. This study presents a novel backward algorithm grounded in the construction of survival functions, specifically designed to tackle the issue of unobservable partial survival times and cure statuses inherent in cure models. We formulate a constrained optimization problem, aiming to maximize the mean survival time of uncured individuals while guaranteeing a predefined cure rate. By leveraging the Lagrange function, we effectively convert this constrained problem into an unconstrained optimization framework, facilitating more tractable analysis. Theoretically, we rigorously derive finite-sample bounds that measure the discrepancy between the estimated treatment regime and the optimal DTR, considering both the cure rate and the average survival time of uncured patients as key performance indicators. Comprehensive numerical simulations demonstrate the method’s superiority, robustness, and versatility across diverse scenarios. Applying our approach to real-world colorectal cancer patient data validates its practical effectiveness and underscores its potential to advance precision medicine treatment strategies.
报告人简介：
林存洁，中国人民大学统计学院副教授，主要研究方向包括生存分析，精准医疗，亚组分析，迁移学习，健康医疗数据建模与分析等，在国际统计期刊《Biometrics》《Annals of Applied Statistics》《Journal of Computational and Graphical Statistics》等发表学术论文30余篇，主持国家自然科学基金项目、国家统计局重点项目等，担任世界中医药学会联合会临床疗效评价专业委员会常务理事；北京生物医学统计与数据管理研究会理事；中国现场统计研究会资源与环境统计分会理事等。