Aselisewine Wisdom, Pal Suvra
Department of Mathematics, University of Texas at Arlington, Arlington, TX 76019, United States.
Division of Data Science, College of Science, University of Texas at Arlington, Arlington, TX 76019, United States.
Stat Comput. 2025 Oct;35(5). doi: 10.1007/s11222-025-10674-y. Epub 2025 Jun 22.
The mixture cure rate model (MCM) is commonly used for analyzing survival data with a cured subgroup. While the prevailing approach to modeling the probability of cure involves a generalized linear model using a known parametric link function, such as the logit link function, it has limitations in capturing the complex effects of covariates on cure probability. This paper introduces a novel MCM employing a neural network-based classifier for cure probability and an accelerated failure time structure for the survival distribution of uncured patients. An expectation maximization algorithm is developed for parameter estimation. Simulation results demonstrate the superior performance of the proposed model in capturing non-linear classification boundaries compared to logit-based and spline-based MCMs, as well as other machine learning algorithms. This enhances the accuracy and precision of cured probability estimates, improving predictive accuracy. The proposed model and estimation method are applied to survival data on leukemia cancer patients, showcasing their effectiveness.
混合治愈率模型（MCM）通常用于分析具有治愈亚组的生存数据。虽然目前对治愈概率建模的方法涉及使用已知参数链接函数（如对数链接函数）的广义线性模型，但它在捕捉协变量对治愈概率的复杂影响方面存在局限性。本文介绍了一种新颖的MCM，它采用基于神经网络的分类器来估计治愈概率，并采用加速失效时间结构来描述未治愈患者的生存分布。开发了一种期望最大化算法进行参数估计。模拟结果表明，与基于对数和样条的MCM以及其他机器学习算法相比，所提出的模型在捕捉非线性分类边界方面具有优越性能。这提高了治愈概率估计的准确性和精度，提升了预测精度。所提出的模型和估计方法应用于白血病癌症患者的生存数据，展示了它们的有效性。
