Assessing the accuracy of automated CT perfusion software in excluding acute stroke: a comparative study of two software packages

TL;DR

This study compares two CTP software packages for detecting stroke and finds one is much better at avoiding false positives.

Contribution

The study reveals significant variability in CTP software accuracy for excluding small lacunar infarcts.

Findings

CMN showed 98.3% specificity with zero infarct volume in 57/58 patients.

Syngo.via settings produced false-positive ischemic cores with volumes up to 207.9 mL.

High specificity in CTP could reduce reliance on follow-up MRI in stroke assessment.

Abstract

Computed tomography perfusion (CTP) is frequently used for the rapid assessment of suspected acute ischemic stroke (AIS). However, small lacunar infarcts often remain undetected by automated software, leading to false negatives and additional imaging. We compared the specificity of two commonly used CTP software packages in patients without evidence of stroke on follow-up diffusion-weighted imaging (DWI).

In this single-center retrospective study, 58 consecutive patients with suspected AIS but negative follow-up DWI–MRI were included. All patients underwent CTP on the same scanner. Perfusion data were processed using (1) syngo.via (Siemens Healthcare) with three parameter settings—A: CBV < 1.2 mL/100 mL, B: additional smoothing filter, and C: rCBF <30%—and (2) Cercare Medical Neurosuite (CMN). Software-reported ischemic core volumes were compared with the MRI findings.

CMN showed the highest specificity, indicating zero infarct volume in 57/58 patients (98.3%). Conversely, all three syngo.via settings produced false-positive ischemic cores, with median volumes ranging from 21.3 mL (setting C) to 92.1 mL (setting A). Only syngo.via setting C reported zero infarct volume in some patients, yet still showed substantial overestimation (maximum 207.9 mL).

Our findings underscore the significant variability in the ability of different CTP software packages to reliably rule out small (lacunar) infarcts. CMN demonstrated good specificity, suggesting that dependable CTP-based stroke exclusion is achievable with advanced post-processing. High specificity could reduce reliance on follow-up MRI in acute stroke pathways if validated, thereby improving resource allocation and patient throughput.

Figures