Early postoperative ipsilateral middle cerebral artery pulsatility index is associated with cerebral hemorrhage after endovascular treatment in anterior circulation large vessel occlusion

Objective To investigate the predictive value of the early postoperative ipsilateral middle cerebral artery pulsatility index (PI) for cerebral hemorrhage after endovascular treatment in patients with anterior circulation large vessel occlusion acute ischemic stroke. Methods A total of 120 patients were divided into a cerebral hemorrhage group ( n = 17) and a non-hemorrhage group ( n = 103). Multivariable logistic regression and Firth penalized likelihood regression were used to identify independent predictors. Receiver operating characteristic (ROC) curves evaluated the predictive performance of individual indicators and a combined model. Bootstrap resampling (1,000 iterations) was performed for internal validation. Results Multivariable logistic regression showed that diabetes mellitus (OR = 4.154, 95% CI: 1.255–13.756, p = 0.020), ASPECTS >8 (OR = 0.115, 95% CI: 0.014–0.942, p = 0.044), and elevated ipsilateral PI (OR = 11.355, 95% CI: 1.267–101.761, p = 0.030) were associated with cerebral hemorrhage. Firth regression confirmed these results. The ipsilateral PI predicted cerebral hemorrhage with an AUC of 0.716 (95% CI: 0.597–0.835), an optimal cutoff of 0.735, a sensitivity of 0.882, and a specificity of 0.505. The combined model (diabetes, ASPECTS >8, PI) achieved an AUC of 0.808 (95% CI: 0.718–0.899), with a sensitivity of 0.882 and a specificity of 0.650. Bootstrap internal validation showed a bias-corrected AUC of 0.810 (95% CI: 0.717–0.896) and an optimism of −0.0022. Conclusion The early postoperative ipsilateral middle cerebral artery PI is a potential prognostic marker for any radiographic intracranial hemorrhage after endovascular treatment in anterior circulation large vessel occlusion acute ischemic stroke, with an optimal cutoff of 0.735. In this cohort, adding PI to diabetes mellitus and ASPECTS was associated with a modest improvement in discrimination (ΔAUC = 0.088, p = 0.011) and continuous net reclassification improvement (NRI = 0.594, p = 0.020), whereas categorical NRI was not significant. These findings are exploratory and require validation in larger cohorts.