Optimizing CTP in Posterior Circulation Infarction (POCI): A Comprehensive Analysis of CTP Postprocessing Algorithms for POCI.

Abstract

Background and purpose: CTP software packages utilize various mathematical techniques to transform source data into clinically useful maps. These techniques have not been validated for posterior circulation infarction (POCI). Studies of anterior circulation stroke have shown that algorithm differences significantly influence the accuracy and best tissue parameters and thresholds of output maps. We examined the influence of the processing algorithm on CTP accuracy and best tissue parameters and thresholds in acute POCI.

Materials and methods: Data were analyzed from patients diagnosed with a POCI enrolled in the International Stroke Perfusion Imaging Registry (INSPIRE). Fifty-eight patients with baseline multimodal CT with occlusion of a large posterior circulation artery and follow-up diffusion-weighted MRI at 24-48 hours were included. CTP parametric maps were generated by using 5 algorithms, including singular value deconvolution, singular value deconvolution with delay and dispersion correction (SVDd), partial deconvolution, the stroke stenosis, and maximum slope models. Receiver operating curve (ROC) analysis and linear regression were used for voxel-based analysis and volume-based analysis, respectively.

Results: Partial deconvolution by using the MTT parameter was the optimal technique for characterizing ischemic-penumbra (AUC 95% CI: 0.73 [0.64-0.81]) and infarct core (AUC 95% CI: 0.70 [0.63-0.73]). The optimal MTT threshold was >165% and >180% for core and penumbra, respectively. By volume analysis, the maximum slope and SVDd by using MTT were the best algorithms for the estimation of penumbra and core, respectively. Estimates of core volume were weak (all R² ≤0.02). Processing algorithm-influenced model accuracy (AUC range: 0.70-0.73 [core], 0.67-0.72 [penumbra]) and optimal tissue parameter and threshold. MTT was the most consistent optimal parameter across algorithms. The optimal MTT threshold by algorithm varied from >120% to >200% for core and 155% to 195% for penumbra.

Conclusions: CTP has diagnostic utility in POCI. There were notable differences in optimal parameter and threshold by algorithm. Clinicians should be aware of the specific characteristics of the algorithm used in their CTP software and apply caution when comparing output maps between vendors.