StrokeTimer: Robust Representation Learning for Ischemic Stroke Onset-Time Estimation from Non-contrast CT

2026-06-03

AI Summary

This study addresses the clinical challenge of uncertain onset time in acute ischemic stroke, which often precludes timely reperfusion therapy. To overcome this limitation, the authors propose StrokeTimer, a novel framework that, for the first time, integrates self-supervised disentangled learning with energy-guided contrastive learning to extract subtle ischemic signs from routine non-contrast CT scans. The method effectively handles class imbalance and data heterogeneity across multicenter cohorts by estimating symptom onset within three clinically relevant time windows. Evaluated on the MR CLEAN Registry and MR CLEAN LATE national cohorts, StrokeTimer achieves a macro AUC of 0.69 and a macro F1-score of 0.57, representing an improvement of nearly 50% over the strongest baseline (p < 0.005) and significantly outperforming existing approaches.

Abstract

Ischemic stroke is a major global disease. Treatment decisions are highly time-sensitive, as eligibility for reperfusion therapies relies on the interval between stroke onset and intervention. However, the true onset time is often uncertain in clinical practice, necessitating imaging-based assessment of tissue age as a surrogate marker. Early ischemic changes on routinely acquired non-contrast CT (NCCT) are often subtle, and real-world clinical datasets exhibit pronounced onset-time class imbalance and center-scanner-related heterogeneity. In this work, we propose StrokeTimer, a fully automated framework for onset-time estimation in acute ischemic stroke. StrokeTimer integrates self-supervised disentanglement learning with energy-guided contrastive learning to capture subtle ischemic patterns while addressing long-tailed data distributions under acquisition variability. Onset time is categorized into three clinically relevant windows: <4.5 h, 4.5-6 h, and >6 h. Experimental results on a large multi-center NCCT dataset from two national cohorts, MR CLEAN Registry and MR CLEAN LATE, show that StrokeTimer achieves a macro AUC of 0.69 and a macro F1-score of 0.57, improving the strongest baseline by nearly 50% (p < 0.005). In this realistic, challenging setting, representative baseline approaches exhibit near-chance macro performance. Model explanations further highlight subtle gray-white matter blurring and hypodense regions consistent with established radiological biomarkers. These findings demonstrate the potential of StrokeTimer to support treatment decision-making in acute ischemic stroke. Code is available at https://github.com/BrainVas/StrokeTimer.

Problem

• ischemic stroke
• onset-time estimation
• non-contrast CT
• class imbalance
• scanner heterogeneity

Innovation

• self-supervised disentanglement learning
• energy-guided contrastive learning
• ischemic stroke onset-time estimation
• non-contrast CT
• long-tailed data distribution