Objectives Computed tomography perfusion (CTP) data are important for hyperacute stroke decision making. Comparisons between outputs of different CTP software packages are limited. We aimed to assess the pair-wise differences in infarct and penumbra estimates produced by three CTP software packages – MIStar, RAPID, and Vitrea.

Methods Consecutive patients with suspected acute ischaemic stroke who underwent CTP between July 2020 and June 2021 at our hospital were independently reviewed by two expert readers. Pair-wise differences between software estimates of penumbra and core volumes were calculated for each patient, with analysis stratified by large vessel occlusion (LVO) status (no-LVO, proximal M2, M1 and internal carotid artery-T [ICA-T]).

Results 580 CTP studies were performed; 262 were normal, 146 technically poor, with 172 included in the final analysis. 79/172 (45.9%) had LVO; proximal M2 (n=21), M1 (n=38) and ICA-T (n=20). Overall, statistically significant pair-wise differences were seen for both penumbra and core estimates (P <0.001). The largest difference in mean core estimates were seen between Vitrea and MIStar ([mean, 95% confidence interval] no-LVO [5.8ml, 3.2–8.4]; proximal M2 [10.4ml, 3.9–17.0]; M1 [17.7ml, 8.9–26.6]; ICA-T [38.9ml, 20.2–57.7]). More comparable penumbra estimates were observed between RAPID and MIStar (no-LVO [1.79ml, -3.9–7.51]; proximal M2 [13.1ml, -0.24–26.5]; M1 [10.7ml, -5.9–27.3]; ICA-T [28.4ml, 0.78–56.0]).

Conclusion Core and penumbra volume estimates vary significantly between CTP software packages. There are minimal differences in patients with non-LVO stroke, with the greatest differences seen in patients with ICA-T occlusions.