CT perfusion based ASPECTS improves the diagnostic performance of early ischemic changes in large vessel occlusion

Tie-Gong Wang, Luguang Chen, Xianglan Jin, Yuan Yuan, Qian-Wen Zhang, Chengwei Shao, Jianping LuBMC Medical Imaging 2021 · DOI

TL;DR

This study demonstrates that CT Perfusion (CTP)-based ASPECTS scores are superior to traditional Non-Contrast CT (NCCT) and automated ASPECTS in identifying early ischemic changes in patients with large vessel occlusion (LVO) stroke. CTP-ASPECTS also proves to be a better predictor of clinical outcomes at 90 days, suggesting its potential to improve patient selection for advanced stroke treatments.

What Problem Does This Paper Address?

Acute ischemic stroke (AIS) due to large vessel occlusion (LVO) is a time-critical medical emergency. Early identification of ischemic changes in the brain is crucial for guiding treatment decisions, particularly for reperfusion therapies like mechanical thrombectomy. The Alberta Stroke Program Early CT Score (ASPECTS) is a widely used method to quantify these early ischemic changes (EICs) on NCCT. However, NCCT-ASPECTS has limitations, including inter-observer variability and difficulty in detecting subtle changes in the early stages. While automated ASPECTS (Auto-ASPECTS) aim to reduce variability, and CT Perfusion (CTP) offers more detailed physiological information, it's unclear which scoring method is most effective for predicting clinical outcomes in LVO patients. This paper aims to compare the diagnostic performance of Auto-, NCCT-, and CTP-ASPECTS for EICs and to determine which method best predicts clinical outcomes in this patient population.

Methods

The study retrospectively analyzed data from 81 patients with anterior circulation LVO who underwent multimodal CT scans (NCCT, CTA, and CTP) within 20 minutes of admission. Patients were categorized into good (modified Rankin Scale [mRS] 0-2) or poor (mRS 3-6) clinical outcomes at 90 days.

Three ASPECTS scoring methods were employed:

  • NCCT-ASPECTS: Standard scoring based on non-contrast CT images, assessed by two blinded radiologists.
  • Auto-ASPECTS: Automated scoring performed by commercial software (RAPID).
  • CTP-ASPECTS: Assessed qualitatively on CTP mismatch maps by a third radiologist, focusing on regions with reduced cerebral blood flow (rCBF < 30%).

The study compared the diagnostic performance of these methods in detecting EICs and predicting clinical outcomes using receiver operating characteristic (ROC) curves, calculating areas under the curve (AUCs), sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). Statistical analyses included Mann-Whitney U tests, chi-squared tests, Friedman's two-way analysis of variance, and Bonferroni correction for post-hoc comparisons.

Datasets

The study utilized a dataset of 81 AIS patients with intracranial LVO treated at Changhai Hospital of Shanghai between January 2019 and June 2019. After exclusions, 63 patients were included in the final analysis. No new benchmark datasets were introduced.

Key Insights and Ideas

  • CTP-ASPECTS Captures More Subtle Ischemia: The CTP-ASPECTS was consistently lower than Auto- and NCCT-ASPECTS, suggesting it is more sensitive in identifying early ischemic changes, potentially reflecting a more accurate representation of the true ischemic territory.
  • Predictive Power of CTP-ASPECTS: CTP-ASPECTS demonstrated significantly higher diagnostic performance (AUCs) for predicting clinical outcomes compared to both Auto- and NCCT-ASPECTS.
  • Automated vs. Manual NCCT: Auto-ASPECTS and NCCT-ASPECTS showed similar performance, indicating that automated systems can effectively replicate manual scoring for NCCT, potentially saving time and reducing inter-observer variability.
  • Clinical Relevance of Dichotomized Scoring: The study found that dichotomizing the ASPECTS (e.g., ASPECTS ≥ 6 vs. < 6) was useful, and CTP-ASPECTS with a cutoff of 6 showed good sensitivity and moderate specificity for predicting good clinical outcomes.

Key Results

  • Diagnostic Performance: The AUC for ordinal CTP-ASPECTS was 0.870, significantly higher than Auto-ASPECTS (0.697) and NCCT-ASPECTS (0.730). Similarly, for dichotomized scoring, CTP-ASPECTS achieved an AUC of 0.829, outperforming Auto-ASPECTS (0.676) and NCCT-ASPECTS (0.667).
  • Predicting Clinical Outcome: CTP-ASPECTS with a cutoff value of 6 was identified as a good predictor of clinical outcome at 90-day follow-up, showing good sensitivity (91.7% for ordinal, 91.7% for dichotomized) and better specificity (74.1% for ordinal, 74.1% for dichotomized) compared to the other methods.
  • Score Differences: Significant differences were observed in ordinal and dichotomized ASPECTS scores between patients with good and poor clinical outcomes across all three methods (p < 0.01), with lower scores generally associated with poorer outcomes.

Limitations and Future Work

The authors acknowledge several limitations:

  • Retrospective and Single-Center Study: The findings may not be generalizable to other populations or healthcare settings.
  • Selection Bias: The study focused only on patients with MCA and IICA occlusion, potentially limiting the applicability to other types of LVO.
  • Incomplete Follow-up Imaging: The absence of follow-up CT or MRI scans in some patients prevented a direct comparison of ASPECTS with final infarct volume.
  • Heterogeneity in Treatment Strategies: While statistical analysis attempted to account for this, the inclusion of patients treated with different reperfusion strategies (antiplatelet therapy, intravenous thrombolysis, mechanical thrombectomy) could introduce variability.

Future work could involve larger, multi-center prospective studies to confirm these findings. The authors also suggest stratifying patients based on treatment strategies (no reperfusion, thrombolysis only, thrombectomy) to further refine the predictive value of CTP-ASPECTS.

Summary updated

BibTeX
@article{wang2021ct,
  author = {Tie-Gong Wang and Luguang Chen and Xianglan Jin and Yuan Yuan and Qian-Wen Zhang and Chengwei Shao and Jianping Lu},
  title = {{CT perfusion based ASPECTS improves the diagnostic performance of early ischemic changes in large vessel occlusion}},
  journal = {BMC Medical Imaging},
  year = {2021},
  doi = {10.1186/s12880-021-00593-5},
}
Figure