arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2509.04485cs.CLcs.AI
收藏

ASCENDgpt:一种基于电子健康记录进行心血管风险预测的表型感知Transformer模型

ASCENDgpt: A Phenotype-Aware Transformer Model for Cardiovascular Risk Prediction from Electronic Health Records

NHS Greater Glasgow and Clyde(英国格拉斯哥和克莱德国家健康服务署);University of Birmingham(伯明翰大学)

Chris Sainsbury, Andreas Karwath

AI总结:

本研究提出用于心血管风险预测的表型感知Transformer模型ASCENDgpt,通过表型感知分词大幅压缩词表,经预训练微调后在五类心血管结局预测上取得优异C-index表现,兼顾临床可解释性与计算效率。

摘要:

我们提出了ASCENDgpt,这是一种专为基于纵向电子健康记录(EHR)进行心血管风险预测而设计的基于transformer的模型。我们的方法引入了一种新颖的表型感知分词方案,将47155个原始ICD编码映射为176个具有临床意义的表型token,在保留语义信息的同时实现了99.6%的诊断编码整合。这种表型映射带来的总词表量为10442个token——与直接使用原始ICD编码相比减少了77.9%。我们在源自19402名独立个体的序列上,以掩码语言建模为目标对ASCENDgpt进行预训练,随后针对五种心血管结局的事件发生时间预测进行微调,这五种结局分别是:心肌梗死(MI)、卒中、主要不良心血管事件(MACE)、心血管死亡以及全因死亡率。我们的模型在留出测试集上展现出优异的区分能力,平均C-index为0.816,在所有结局上均表现出色(MI:0.792,卒中:0.824,MACE:0.800,心血管死亡:0.842,全因死亡率:0.824)。这种基于表型的方法在保证计算效率的同时,还能实现具有临床可解释性的预测。我们的工作证明了领域特定分词和预训练在基于EHR的风险预测任务中的有效性。

英文摘要:

We present ASCENDgpt, a transformer-based model specifically designed for cardiovascular risk prediction from longitudinal electronic health records (EHRs). Our approach introduces a novel phenotype-aware tokenization scheme that maps 47,155 raw ICD codes to 176 clinically meaningful phenotype tokens, achieving 99.6\% consolidation of diagnosis codes while preserving semantic information. This phenotype mapping contributes to a total vocabulary of 10,442 tokens - a 77.9\% reduction when compared with using raw ICD codes directly. We pretrain ASCENDgpt on sequences derived from 19402 unique individuals using a masked language modeling objective, then fine-tune for time-to-event prediction of five cardiovascular outcomes: myocardial infarction (MI), stroke, major adverse cardiovascular events (MACE), cardiovascular death, and all-cause mortality. Our model achieves excellent discrimination on the held-out test set with an average C-index of 0.816, demonstrating strong performance across all outcomes (MI: 0.792, stroke: 0.824, MACE: 0.800, cardiovascular death: 0.842, all-cause mortality: 0.824). The phenotype-based approach enables clinically interpretable predictions while maintaining computational efficiency. Our work demonstrates the effectiveness of domain-specific tokenization and pretraining for EHR-based risk prediction tasks.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑