A Comparison of Three Perfusion Algorithms in Patients at Risk of Delayed Cerebral Ischemia After Subarachnoid Hemorrhage.

Background/

Objectives

Delayed cerebral ischemia (DCI) after an aneurysmal subarachnoid hemorrhage (aSAH) often presents with bilateral vasospasm and cortical spreading depolarizations. Computer tomography perfusion (CTP) is the prevailing screening method for detecting early changes in the cerebral blood flow. Commonly used CTP thresholds include an rCBF < 30% for the core volume and a Tmax > 6 s for hypoperfused tissue detection in acute ischemic stroke. These stroke algorithm computing thresholds compared to the contralateral hemisphere may or may not apply to detect tissue at risk of DCI. We aimed to quantify the volumetric agreement of three different stroke algorithms compared to the final infarct volumes as the standard.

Methods

Furthermore, 123 CTP datasets of 75 patients with aSAH suspicious of DCI were processed using Intellispace Portal (ISP), Cercare Threshold, and Cercare Artificial Intelligence (AI) to calculate the tissue-at-risk (hypoperfused) and non-viable tissue (core) volumes. CT infarct volumes in plain CTs were segmented in the follow-up study by using a 3D slicer.

Results

The calculated core volumes corresponded best to the final infarct volumes if DCI-related treatment was performed subsequently. Additional postprocessing improved the calculation of core volumes but overestimated the tissue at risk of hypoperfusion in DCI. Whereas the accuracy of tissue-at-risk prediction accelerated without treatment, underlining the importance of intra-arterial spasmolysis and induced hypertension in the prevention of DCI.

Conclusions

Cercare AI and ISP revealed a sensitivity of 100% each, with a serious low specificity of <5% that was independent of treatment. Overall, the Cercare Threshold, applying the commonly used stroke thresholds, performed the best in predicting tissue at risk of hypoperfusion in DCI.