Document resource
Introduction Cerebral vasospasm (CVS) remains a major complication of aneurysmal subarachnoid hemorrhage (aSAH), affecting up to 50% of patients. Prior work from our group demonstrated the feasibility of combining transcranial Doppler (TCD) and computed tomography angiography (CTA) to model flow within the Circle of Willis (CoW), including the use of 1-dimensional (1D) modeling with results comparable to 3-dimensional (3D) simulations. We developed a fully automated, open-source pipeline for quantifying cerebral blood flow in the CoW directly from CTA and TCD data at presentation. This pipeline enables automatic generation of labeled centerline vascular networks from CTA, eliminating the need for specialized expertise. We evaluated whether this automated approach could identify baseline hemodynamic differences predictive of CVS.Methods Fifteen patients were included (n=5 per group): (1) aSAH with CVS, (2) aSAH without CVS, and (3) perimesencephalic hemorrhage as a control. CTA data underwent automated segmentation and vessel labeling to generate patient-specific CoW centerline networks. TCD waveforms were extracted using an automated algorithm. Bayesian inference was then applied to generate patient-specific 1D hemodynamic models, enabling estimation of vessel-level resistance and capacitance at presentation. These parameters were compared across groups.Results The fully automated pipeline significantly reduced processing time for vessel segmentation and TCD data extraction while maintaining results comparable to prior manually generated 1D models. Preliminary analysis demonstrates differences in baseline resistance and capacitance in patients who developed CVS compared to CVS-negative and control groups.Conclusion By fully automating and open-sourcing a pipeline for quantifying cerebral blood flow in the CoW, we provide a clinically practical tool for early hemodynamic assessment in aSAH. This approach eliminates the need for specialized modeling expertise and enables scalable, patient-specific analysis at presentation. Automated 1D modeling may facilitate early risk stratification for CVS and help identify patients who would benefit from closer monitoring or early intervention.Disclosures Z. Abecassis: None. M. Fung: None. J. Lim: None.