Automated Net Water Uptake Assessment in Acute Ischemic Stroke Using NCCT
Overview
This study presents a fully automated, explainable method to quantify net water uptake (NWU) and infarct volume from non-contrast CT (NCCT) images alone in acute ischemic stroke (AIS). Validation on two independent datasets demonstrated that NWU maps derived solely from NCCT correlate well with expert-annotated lesion masks from CT perfusion (CTP) and diffusion-weighted imaging (DWI), offering a practical alternative to conventional multi-modality imaging.
Background
Infarct lesion volume is a critical endpoint in AIS clinical trials, traditionally assessed by follow-up imaging or CT perfusion, which has limitations including availability and potential overestimation of irreversible damage. NWU, calculated from NCCT hypodensity, is a promising prognostic biomarker but typically requires manual lesion delineation using CTP or DWI. Automated NCCT-only approaches could reduce radiation exposure, streamline workflows, and minimize treatment delays. However, infarct segmentation on NCCT is challenging due to subtle hypodensity and variability in image quality.
Data Highlights
Dataset
Patients Included
Imaging Modalities
Lesion Reference
Exclusions
In-house (Hamburg)
155
NCCT, CTP
CTP-derived TTD masks
30 excluded (18 poor quality, 12 small lesions)
External (Boston)
46
NCCT, DWI/ADC
DWI lesion masks
5 excluded (poor quality)
Key Findings
The automated pipeline accurately estimated NWU and infarct volumes from NCCT images without requiring CTP or DWI input.
NWU maps showed strong agreement with expert-annotated lesion masks derived from CTP in the in-house dataset.
External validation confirmed the method's robustness by correlating NWU maps with DWI-based lesion masks.
The approach avoids deep learning, using interpretable, data-efficient heuristics, reducing overfitting risk and enhancing reproducibility.
Automated NWU quantification from NCCT alone may help distinguish reversible from irreversible ischemic tissue damage.
Clinical Implications
This automated NCCT-based NWU quantification method offers a radiation-sparing, widely accessible tool for early ischemic stroke assessment, potentially improving patient selection for reperfusion therapies. By reducing reliance on advanced imaging modalities and manual segmentation, it may streamline clinical workflows and expedite treatment decisions. The approach's transparency and reproducibility support its integration into routine stroke imaging protocols.
Conclusion
The study demonstrates that fully automated NWU estimation from NCCT images alone is feasible and correlates well with established imaging references, providing a practical alternative to conventional multi-modality stroke imaging. This advancement may enhance acute stroke evaluation and treatment planning in diverse clinical settings.
References
Kumar et al 2024 -- Automated NWU quantification in AIS
Nowinski et al 2023 -- Stroke imaging marker-based techniques
