Saved in:
Bibliographic Details
Main Authors: Zheng, Zhichong, Nie, Xiaohang, Wang, Xueqi, Zhao, Yuanjin, Zhang, Haitao, Tang, Yichao
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2604.01727
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910097074028544
author Zheng, Zhichong
Nie, Xiaohang
Wang, Xueqi
Zhao, Yuanjin
Zhang, Haitao
Tang, Yichao
author_facet Zheng, Zhichong
Nie, Xiaohang
Wang, Xueqi
Zhao, Yuanjin
Zhang, Haitao
Tang, Yichao
contents Forecasting evolving clinical risks relies on intrinsic pathological dependencies rather than mere chronological proximity, yet current methods struggle with coarse binary supervision and physical timestamps. To align predictive modeling with clinical logic, we propose the Medical-semantics Aware Time-ALiBi Transformer (MATA-Former), utilizing event semantics to dynamically parameterize attention weights to prioritize causal validity over time lags. Furthermore, we introduce Plateau-Gaussian Soft Labeling (PSL), reformulating binary classification into continuous multi-horizon regression for full-trajectory risk modeling. Evaluated on SIICU -- a newly constructed dataset featuring over 506k events with rigorous expert-verified, fine-grained annotations -- and the MIMIC-IV dataset, our framework demonstrates superior efficacy and robust generalization in capturing risks from text-intensive, irregular clinical time series.
format Preprint
id arxiv_https___arxiv_org_abs_2604_01727
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle MATA-Former & SIICU: Semantic Aware Temporal Alignment for High-Fidelity ICU Risk Prediction
Zheng, Zhichong
Nie, Xiaohang
Wang, Xueqi
Zhao, Yuanjin
Zhang, Haitao
Tang, Yichao
Machine Learning
Forecasting evolving clinical risks relies on intrinsic pathological dependencies rather than mere chronological proximity, yet current methods struggle with coarse binary supervision and physical timestamps. To align predictive modeling with clinical logic, we propose the Medical-semantics Aware Time-ALiBi Transformer (MATA-Former), utilizing event semantics to dynamically parameterize attention weights to prioritize causal validity over time lags. Furthermore, we introduce Plateau-Gaussian Soft Labeling (PSL), reformulating binary classification into continuous multi-horizon regression for full-trajectory risk modeling. Evaluated on SIICU -- a newly constructed dataset featuring over 506k events with rigorous expert-verified, fine-grained annotations -- and the MIMIC-IV dataset, our framework demonstrates superior efficacy and robust generalization in capturing risks from text-intensive, irregular clinical time series.
title MATA-Former & SIICU: Semantic Aware Temporal Alignment for High-Fidelity ICU Risk Prediction
topic Machine Learning
url https://arxiv.org/abs/2604.01727