Acute infarction after mechanical thrombectomy is better delineable in virtual non-contrast compared to conventional images using a dual-layer spectral CT

Isabelle Riederer, Alexander A. Fingerle, Thomas Baum, Jan S. Kirschke, Ernst J. Rummeny, Peter B. Noël, Daniela PfeifferScientific 2018 · DOI

Abstract

Following mechanical thrombectomy for acute ischemic stroke, iodinated contrast extravasation often complicates post-procedural imaging by masking infarcts or mimicking hemorrhages. This study evaluates the efficacy of Virtual Non-Contrast (VNC) images, derived from dual-layer spectral CT, in improving the detection and volumetric assessment of acute infarcts compared to conventional CT. In a study of 29 patients, VNC-CT demonstrated significantly higher contrast-to-noise ratios (3.1 vs 1.1), improved sensitivity (73% vs 55%), and more accurate volume estimation (72% vs 55% of MRI volume) compared to conventional imaging.

Introduction

Mechanical thrombectomy has revolutionized the treatment of large vessel occlusion strokes. However, the use of iodinated contrast during the procedure frequently leads to extravasation into brain tissue, particularly when the blood-brain barrier is compromised. On conventional CT scans, this leaked contrast appears hyperdense (bright), which can obscure the hypodense (dark) signal of the underlying infarct or be mistaken for an intracranial hemorrhage. While MRI remains the gold standard for assessing final infarct volume, its limited availability and longer scan times make it less ideal for immediate post-operative monitoring. There is a critical clinical need for a CT-based method that can reliably differentiate between contrast agent and ischemic tissue.

Methods

The researchers conducted a retrospective analysis of 29 patients who underwent unenhanced head spectral CT approximately 12 hours after mechanical thrombectomy. The imaging was performed using a dual-layer spectral CT system, which captures low- and high-energy photon data simultaneously without increasing radiation dose. This spectral data allows for the generation of Virtual Non-Contrast (VNC) images by mathematically "subtracting" the iodine signal from the scan.

To validate the findings, follow-up imaging (primarily Diffusion-Weighted MRI) was obtained an average of 2.4 days later to serve as the ground truth for infarct volume. Two independent readers assessed the images using the Alberta Stroke Program Early CT Score (ASPECTS). Key metrics included the Contrast-to-Noise Ratio (CNR), sensitivity, specificity, and semi-automated volume measurements.

Results

The visual clarity provided by VNC-CT was significantly superior to conventional imaging. By removing the "fog" of iodinated contrast, the underlying ischemic lesions became much more apparent, matching more closely with the gold-standard MRI results.

Comparison of conventional CT, VNC-CT, and gold-standard DWI-MRI across three clinical cases. Arrows indicate areas of infarction and contrast extravasation.

Quantitative analysis supported these visual observations across several key performance indicators:

  • Contrast-to-Noise Ratio (CNR): VNC-CT achieved a CNR of 3.1 ± 1.5, nearly triple the 1.1 ± 1.1 observed in conventional CT (p < 0.001).
  • Sensitivity: VNC-CT demonstrated a sensitivity of 73% for identifying ischemic lesions, compared to just 55% for conventional CT.
  • Volume Accuracy: VNC-CT estimated infarct volume at 72% of the MRI-measured volume, whereas conventional CT only captured 55% (p < 0.005).
  • Inter-rater Reliability: Agreement between the two readers was higher for VNC-CT (0.751) than for conventional CT (0.625).
Sensitivity and specificity metrics for two independent readers comparing conventional CT and VNC-CT.

Discussion and Conclusion

Virtual Non-Contrast imaging using dual-layer spectral CT significantly enhances the diagnostic accuracy of post-thrombectomy brain scans. By effectively isolating ischemic tissue from extravasated contrast, VNC-CT provides clinicians with a clearer understanding of stroke extent and potential complications without requiring additional radiation or specialized protocols.

While the study is limited by its small sample size and the temporal gap between CT and follow-up MRI, the results suggest that VNC-CT could become a standard tool in the post-operative management of stroke patients. Future advancements, such as photon-counting CT, may further refine these diagnostic capabilities and improve sensitivity in the early post-operative window.

Summary updated

Abstract

The aim was to evaluate Virtual Non-Contrast (VNC)-CT images for the detection of acute infarcts in the brain after mechanical thrombectomy using a dual-layer spectral CT. 29 patients between September 2016 and February 2017 with unenhanced head spectral-CT after mechanical thrombectomy and available follow-up images (MRI, n:26; CT, n:3) were included. VNC-CT and conventional CT (CT) images were reconstructed using dedicated software. Based on those, contrast-to-noise ratio (CNR), and the volume of infarction were measured semi-automatically in VNC-CT, CT and MRI. Furthermore, two readers independently assessed the VNC-CT and CT images in a randomized order by using the ASPECT score, and inter-rater reliability, sensitivity and specificity were calculated. CNR was significantly higher in VNC-CT compared to CT (3.1 ± 1.5 versus 1.1 ± 1.1, p < 0.001). The mean estimated volume of infarction was significantly higher in VNC-CT compared to CT (72% versus 55% of the volume measured in MRI, p < 0.005). Inter-rater reliability was higher in VNC-CT compared to CT (0.751 versus 0.625) and sensitivity was higher in VNC-CT compared to CT (73% versus 55%). In conclusion, acute ischemic lesions after mechanical thrombectomy are better definable in VNC-CT compared to CT images using a dual-layer spectral CT system.
BibTeX
@inproceedings{riederer2018acute,
  author = {Isabelle Riederer and Alexander A. Fingerle and Thomas Baum and Jan S. Kirschke and Ernst J. Rummeny and Peter B. Noël and Daniela Pfeiffer},
  title = {{Acute infarction after mechanical thrombectomy is better delineable in virtual non-contrast compared to conventional images using a dual-layer spectral CT}},
  booktitle = {Scientific},
  year = {2018},
  doi = {10.1038/s41598-018-27437-7},
}
Figure