Introduction: Delayed cerebral ischemia (DCI) significantly influences the prognosis of patients with aneurysmal subarachnoid hemorrhage (aSAH), necessitating its early prediction. While pre-treatment computed tomography perfusion (CTP) has shown promise in predicting DCI, its clinical utility has been hampered by challenges in prior studies, including limited sample sizes, inconsistent parameter selection, and the absence of automated, comprehensive whole-brain perfusion processing methods.
Methods: This large, single-center case-control study was derived from a prospectively maintained registry database, Multimodal Computed Tomography in Patients with Acute Hemorrhagic Stroke (MCTAHS), and included patients enrolled between January 2020 and August 2023. Employing the Swin-UNet deep learning model, we generated CTP parameters across 31 discrete brain regions within 2 minutes. The mean values of perfusion parameters were used for statistical analysis. The primary outcome was the presence of DCI, and poor 90-day functional outcomes was the secondary outcome.
Results: Out of 1334 patients, 689 early-admitted individuals who underwent CTP before aneurysm treatment, received emergency aneurysm treatment, and completed follow-up were included. Among patients, 61.2% (422) were female, with an average age of 55.6 ± 11.8 years. Of the parameters, time-to-maximum (Tmax) showed the strongest and most consistent correlation with DCI (Cox regression: hazard ratio 1.39, 95% confidence interval [CI] 1.15–1.70) and poor 90-day functional outcomes (logistic regression: overall risk 2.00, 95% CI 1.46–2.78). Tmax also had the highest area under the curve (AUC) in receiver operating characteristic analysis for predicting DCI (AUC 0.720, 95% CI 0.674–0.767) and poor 90-day functional outcomes (AUC 0.743, 95% CI 0.695–0.791). Subgroup analyses confirmed Tmax’s consistent association with the outcomes across different groups, with no significant intergroup interaction.
Conclusion : Tmax was identified as the most reliable perfusion parameter for predicting DCI. Meanwhile, the post-processing method used for generating CTP parameters in this study was suited to the urgent clinical needs of aSAH.
.jpg)
