Influence of slice thickness on computed tomography perfusion-derived ischemic core volume accuracy in acute ischemic...

Computed tomography (CT) perfusion enables quantitative assessment of cerebral hemodynamic and physiologic parameters and is widely employed for triaging patients with acute ischemic stroke; however, the absence of standardized acquisition protocols for ischemic core estimation remains a major limitation. This study aimed to determine the optimal CT slice thickness for quantifying ischemic core volume derived from CT perfusion in patients with acute ischemic stroke, using diffusion-weighted imaging (DWI)-derived ischemic core volume as the reference standard.

In this retrospective study, patients with acute ischemic stroke who met the inclusion criteria were reviewed. CT perfusion data were reconstructed with slice thicknesses of 1, 3, 5, 7, and 10 mm to evaluate their impact on ischemic core volume quantification. Ischemic core volumes were calculated using fully automated post-processing software. The ischemic core was defined using thresholds of relative cerebral blood flow (rCBF) <20-38% and relative cerebral blood volume (rCBV) <34-42%. The optimal slice thickness was identified based on the lowest mean absolute error (MAE), the smallest bias and narrowest 95% limits of agreement (LoA) in Bland-Altman analysis, and the highest intraclass correlation coefficient (ICC) compared with DWI-derived ischemic core volume.

A total of 87 patients (mean age ± standard deviation, 69.64±12.27 years; 49 males) were included. Among the different slice thicknesses, the lowest MAE was observed at the rCBF <38% threshold compared to rCBF <30%. Bland-Altman analysis and ICC results indicated that 3 mm (bias, 5.04 mL; 95% LoA, -35.29 to 45.38; ICC =0.961), 5 mm (bias, 4.55 mL; 95% LoA, -37.41 to 46.51; ICC =0.961), and 7 mm (bias, 4.75 mL; 95% LoA, -35.14 to 44.66; ICC =0.961) slice thicknesses yielded the most accurate estimates of ischemic core volume when using rCBF <38%. In contrast, 1 mm (bias, 6.71 mL; 95% LoA, -36.86 to 50.29; ICC =0.951) and 10 mm (bias, 9.12 mL; 95% LoA, -47.42 to 65.67; ICC =0.905) reconstructions demonstrated lower agreement with DWI-derived measurements.

CT perfusion data reconstructed at slice thicknesses of 3-7 mm provided the most accurate estimation of ischemic core volume, at both commonly used and optimal perfusion thresholds, compared with thinner (1 mm) or thicker (10 mm) slice thicknesses.