Comparison of CT-perfusion software packages and validation with true infarct core in patients with acute ischemic stroke

Background Computed tomography perfusion (CTP) is widely used to evaluate acute ischemic stroke (AIS) and guide endovascular thrombectomy (MT). However, substantial variability exists among software solutions in estimating ischemic core volume, potentially affecting patient selection and treatment outcomes. Materials and methods We retrospectively analyzed 65 patients with M1/M2 occlusions who underwent successful MT (TICI 2c/3). All patients had baseline CTP and follow-up non-contrast CT (ncCCT) within 10–25 h. Final infarct volumes were segmented on ncCCT. CTP data were processed using syngo.via (Siemens Healthcare) with three settings (A: CBV < 1.2 ml/100 ml; B: same threshold plus smoothing; C: rCBF < 30%) and Cercare Medical Neurosuite (CMN, Cercare Medical). We compared software-derived core volumes to ncCCT -measured volumes using Wilcoxon-Signed-Rank tests, Bland–Altman analyses, and intraclass correlation coefficients. Results CMN showed the smallest median error (0.0 ml) and a 50.0% rate of overestimation. Among syngo.via settings, method B had improved agreement but still tended to overestimate infarct volume in larger strokes. Settings A and C more frequently produced substantial overestimations. Bland–Altman plots demonstrated that deviations from true infarct volumes increased with larger cores for all software packages, underscoring the challenge of accurately quantifying extensive ischemic lesions. Conclusion Our findings reveal marked variability in core volume estimates across different CTP software solutions. CMN and syngo.via B provided good accuracy, but performance declined with larger infarcts. Awareness of these discrepancies is critical for clinicians interpreting perfusion maps to optimize AIS treatment decisions and avoid misclassification of patients who might still benefit from reperfusion therapy.