Diagnostic performance of different perfusion algorithms for the detection of angiographical spasm

To assess the diagnostic utility of different perfusion algorithms for the detection of angiographical terial spasm.

During a 2-year period, 45 datasets from 29 patients (54.2±10,75y, 20F) with suspected cerebral vasospasm after aneurysmal subarachnoid hemorrhage were included. Volume Perfusion CT (VPCT), Non-enhanced CT (NCT) and angiography were performed within 6hours post-ictus. Perfusion maps were generated using a maximum slope (MS) and a deconvolution-based approach (DC). Two blinded neuroradiologists independently evaluated MS and DC maps regarding vasospasm-related perfusion impairment on a 3-point Likert-scale (0=no impairment, 1=impairment affecting <50%, 2=impairment affecting >50% of vascular territory). A third independent neuroradiologist assessed angiography for presence and severity of arterial narrowing on a 3-point Likert scale (0=no narrowing, 1=narrowing affecting <50%, 2=narrowing affecting>50% of artery diameter). MS and DC perfusion maps were evaluated regarding diagnostic accuracy for angiographical arterial spasm with angiography as reference standard. Correlation analysis of angiography findings with both MS and DC perfusion maps was additionally performed. Furthermor, the agreement between MS and DC and inter-reader agreement was assessed.

DC maps yielded significantly higher diagnostic accuracy than MS perfusion maps (DC:AUC=.870; MS:AUC=.805; P=0.007) with higher sensitivity for DC compared to MS (DC:sensitivity=.758; MS:sensitivity=.625). DC maps revealed stronger correlation with angiography than MS (DC: R=.788; MS: R=694;=<0.001). MS and DC showed substantial agreement (Kappa=.626). Regarding inter-reader analysis, (almost) perfect inter-reader agreement was observed for both MS and DC maps (Kappa≥981).

DC yields significantly higher diagnostic accuracy for the detection of angiographic arterial spasm and higher correlation with angiographic findings compared to MS.