CTP-Free Method for Automated Lesion Water Uptake in Acute Ischemic Stroke

BACKGROUND AND PURPOSE: Net water uptake (NWU) in the infarct core of patients with ischemic stroke has been correlated with clinical outcome and lesion age, which could aid in treatment selection. Traditional NWU measurement requires CTP, limiting its clinical applicability. We aimed to develop and evaluate an automated method to measure NWU using only NCCT and CTA. MATERIALS AND METHODS: We included 90 patients with ischemic stroke with known onset time and available NCCT, CTA, and CTP from the Multicenter Randomized Clinical Trial of Endovascular Treatment for Acute Ischemic Stroke in the Netherlands (MR CLEAN) Registry and MR CLEAN-LATE trial. Using deep learning, we automatically segmented the infarct core and hypoperfused area from NCCT and CTA images. NWU was calculated as the relative difference in density between these affected regions and their contralateral counterparts. We included the hypoperfused area because it represents potentially salvageable tissue, and additional NWU analyses in this region could provide insights into ischemic injury progression. We compared this automated CTA-NCCT-based approach with the traditional CTP-NCCT-based approach by assessing their agreement (intraclass correlation coefficient [ICC], Bland-Altman analysis) and accuracy in identifying patients within 4.5 hours of stroke onset (receiver operating characteristic analysis, DeLong test for areas under the curve [AUC] comparison). RESULTS: NWU measured in the core (CTP-NCCT-based: median 4.1%, interquartile range [2.7-6.6]; CTA-NCCT-based: 3.2%, [2.1-5.2]) showed good agreement between approaches (ICC 0.81, 95% CI, 0.73-0.87; mean difference 0.43% [-4.6% to +5.5%]). NWU in the hypoperfused area (CTP-NCCT-based: 2.3%, [1.3-4.1]; CTA-NCCT-based: 2.4%, [0.9-3.9]) showed excellent agreement (ICC 0.93, 95% CI, 0.90-0.96; mean difference 0.17%, -1.54% to +1.88%). For core-based NWU, both approaches detected significantly lower values in patients within versus beyond 4.5 hours (CTP-NCCT-based: 3.7% versus 10%; P < .001; CTA-NCCT-based: 3.1% versus 11%; P < .001) with similar accuracy (AUC, 0.87; P = .88). For hypoperfused area-based NWU, neither approach showed significant differences between patients within versus beyond 4.5 hours (CTP-NCCT-based: 2.3% versus 4.4%; P = .31; CTA-NCCT-based: 2.3% versus 4.7%; P = 0.13) and both had lower accuracy than core-based NWU classification (AUC,: CTP-NCCT-based 0.59, CTA-NCCT-based 0.63; P = 0.81). CONCLUSIONS: The automated CTA-NCCT-based approach shows good agreement with the traditional CTP-NCCT-based method for NWU measurement and achieves similar accuracy in identifying patients within 4.5 hours of onset. External validation is needed to confirm these findings.