Harmonizing CT Perfusion Imaging for Stroke: Standardizing Ischemic Area Estimation

Overview

This study evaluates the variability in CT perfusion (CTP) imaging protocols across multiple stroke centers and their impact on ischemic area estimation. Using an anthropomorphic digital phantom and vendor-specific software, the research highlights significant differences in perfusion parameters and infarct estimations, underscoring the need for standardized imaging protocols.

Background

CT perfusion imaging is critical in acute ischemic stroke evaluation, guiding treatment decisions such as endovascular thrombectomy. However, multicenter studies face challenges due to heterogeneous scan protocols and software algorithms, which affect the consistency of ischemic core and penumbra estimation. Variations in acquisition parameters like tube voltage, exposure, and frame timing, as well as differences in perfusion software preprocessing and algorithms, contribute to inconsistent results. Addressing these discrepancies is essential for reliable clinical guidelines and multicenter research.

Data Highlights

The study collected 18 scan protocols from stroke centers participating in the CLEOPATRA healthcare evaluation, involving 1164 patients from multiple trials. Each protocol's tube voltage (kVp), exposure (mAs), and frame timing were input into a digital anthropomorphic phantom simulating acute ischemic stroke with a ground truth infarct core of 30 mL and penumbra of 55 mL. Ten noise realizations per protocol were generated to assess variability. Perfusion maps were analyzed using three vendor software packages (Philips IntelliSpace Portal, Siemens syngoVIA, and Vital Images Vitrea), each producing different perfusion parameters (CBF, CBV, MTT, TTP/TMAX) with varying export formats and resolutions.

Key Findings

• Significant variability exists in CTP imaging protocols across stroke centers, particularly in tube voltage, exposure, and frame timing.
• Different vendor software packages produce varying perfusion parameter maps and ischemic region estimations due to distinct algorithms and default thresholds.
• The anthropomorphic phantom allowed controlled simulation of ischemic stroke, revealing that noise and protocol differences impact perfusion measurements and infarct volume estimations.
• Vendor software export formats differ, with some providing direct parameter values and others requiring rescaling from grayscale intensities, complicating cross-center data harmonization.
• Adjustments for unavailable scan data (e.g., 70 kVp) were made using established dose equivalence rules, highlighting challenges in standardizing protocols.

Clinical Implications

Clinicians should be aware that variability in CT perfusion acquisition and analysis can lead to inconsistent ischemic core and penumbra estimations, potentially affecting treatment decisions. Standardizing scan protocols and harmonizing vendor software outputs are crucial steps to improve the reliability of multicenter stroke imaging studies and ensure consistent patient evaluation. Adoption of uniform thresholds and preprocessing methods may enhance comparability and clinical utility of CTP imaging.

Conclusion

This study demonstrates that substantial variation in CT perfusion imaging protocols and software analysis exists across stroke centers, impacting ischemic area estimation. Harmonization efforts, including standardized acquisition parameters and vendor software calibration, are essential to improve consistency and validity in stroke imaging research and clinical practice.

References

1. CLEOPATRA Healthcare Evaluation and CONTRAST Consortium Studies
2. Anthropomorphic Digital Phantom for CTP Simulation
3. Vendor Software Algorithms and Perfusion Parameter Differences