CSRA: Controlled Spectral Residual Augmentation for Robust Sepsis Prediction

2026-04-15

AI Summary

This work addresses the performance bottleneck in sepsis prediction under short observation windows, where limited historical data and sparse future labels hinder model efficacy. To overcome this, the authors propose the CSRA framework, which introduces controllable spectral residual augmentation—a novel approach that leverages clinically grouped variables to learn multi-level representations and generates input-adaptive, plausible trajectory variants in the spectral domain. The framework incorporates anchor consistency loss and controller regularization to enable structured, clinically interpretable temporal data augmentation. Trained end-to-end, CSRA achieves a 10.2% reduction in MSE and a 3.7% reduction in MAE on MIMIC-IV, with concurrent improvements in classification performance. It demonstrates robustness across challenging scenarios, including short observation windows, long prediction horizons, limited training samples, and an external dataset (ZiGongICUinfection).

Technology Category

• Machine Learning: Semi-Supervised Learning
• Computer Vision: Multi-modal Vision
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Accurate prediction of future risk and disease progression in sepsis is clinically important for early warning and timely intervention in intensive care. However, short-window sepsis prediction remains challenging, because shorter observation windows provide limited historical evidence, whereas longer prediction horizons reduce the number of patient trajectories with valid future supervision. To address this problem, we propose CSRA, a Controlled Spectral Residual Augmentation framework for short-window multi-system ICU time series. CSRA first groups variables by clinical systems and extracts system-level and global representations. It then performs input-adaptive residual perturbation in the spectral domain to generate structured and clinically plausible trajectory variations. To improve augmentation stability and controllability, CSRA is trained end-to-end with the downstream predictor under a unified objective, together with anchor consistency loss and controller regularization. Experiments on a MIMIC-IV sepsis cohort across multiple downstream models show that CSRA is consistently competitive and often superior, reducing regression error by 10.2\% in MSE and 3.7\% in MAE over the non-augmentation baseline, while also yielding consistent gains on classification. CSRA further maintains more favorable performance under shorter observation windows, longer prediction horizons, and smaller training data scales, while also remaining effective on an external clinical dataset~(ZiGongICUinfection), indicating stronger robustness and generalizability in clinically constrained settings.

Problem

• sepsis prediction
• short-window prediction
• time series augmentation
• robustness
• ICU

Innovation

• Spectral Residual Augmentation
• Short-window Prediction
• ICU Time Series
• Controlled Data Augmentation
• Sepsis Risk Prediction