Article
Automated grading of cerebral vasospasm to standardise computed tomography angiography examinations after subarachnoid haemorrhage
Automatisierte Beurteilung zerebraler Vasospasmen zur Standardisierung von CT-A Untersuchungen nach Subarachnoidalblutung
Search Medline for
Authors
Published: | June 26, 2020 |
---|
Outline
Text
Objective: Computed tomography perfusion imaging (PCT) and computed tomography angiography (CTA) are common diagnostic tools to evaluate the indication for endovascular vasospasm treatment in subarachnoid hemorrhage (SAH) patients with suspected delayed cerebral ischemia (DCI). However, objective parameters for CTA evaluation are lacking. In this study we applied a novel digital and automated CTA evaluation method for investigator-independent detection of vasospasms. We used the method to investigate
- (i.) the link between vasospasm and cerebral perfusion, and
- (ii.) the method’s ability to predict subsequent endovascular vasospasm treatment in a cohort of SAH patients.
Methods: A retrospective chart review and analysis of imaging data of 40 consecutive SAH patients was performed. The cerebrovascular trees were reconstructed three-dimensionally from CTA data and the vessel volume per vessel length ratio (VLR, [µl/mm]) of the arteries of the circle of Willis and their peripheral branches were calculated. Thresholds for the VLR indicating severe vasospasm were determined for each vessel segment after ROC curve analysis based on comparison with digital subtraction angiographies (DSA).
Results: The threshold-based volumetric evaluation of CTA data allowed an automated, investigator-independent detection of severe vasospasms (e.g. VLR < 5.0 µl/mm for the M1 segment). Presence of vasospasm on CTA based on low VLR exhibited a high sensitivity and negative predictive value, but a rather low specificity and positive predictive value to predict cerebral hypoperfusion on PCT (N=96, sensitivity 89%, specificity 41%, NPV 90%, PPV 39% if ≥ 1 vessel segment was affected). The combination of low VLR on CTA and cerebral hypoperfusion on PCT was superior to PCT or low VLR alone in predicting endovascular vasospasm treatment within 24 h after the exam (N=96, sensitivity 58%, specificity 92%, NPV 90%, PPV 65% if ≥ 3 vessel segments were affected).
Conclusion: Digital, volumetric analysis of the cerebrovascular tree allows an objective, investigator-independent detection and quantification of vasospasms. The method could be used as an additional tool in multimodal diagnostics to standardize selection of SAH patients with DCI for endovascular therapies.