Scholarlink

Objective <p>To explore the differences and the prognostic predicted value of three commercial CT perfusion (CTP) post-processing software packages (Packages A, B, and C) in quantifying ischemic core /penumbra and predicting prognoses in acute ischemic stroke (AIS) patients.</p> Methods <p>This is a retrospective analysis of 75 AIS patients processed with three software packages. Agreement was assessed via Bland–Altman plots and intraclass correlation coefficients (ICC). Multivariable regression models evaluated associations between CTP metrics and final infarct volume (FIV), hemorrhagic transformation (HT), and functional outcomes (modified Rankin Scale [mRS]). For model validation, we report the adjusted R<sup>2</sup> for linear models and the area under the curve (AUC) with 95% confidence intervals for logistic models. No internal validation was performed, which is acknowledged as a limitation.</p> Results <p>AI-augmented packages A and B showed higher inter-package agreement and stronger correlation with FIV (r<sub>A</sub> = 0.658, r<sub>B</sub> = 0.675 vs. r<sub>C</sub> = 0.430) and better predictive accuracy for FIV (adjusted R<sup>2</sup>: <i>A</i> = 0.407, <i>B</i> = 0.398 vs. <i>C</i> = 0.199). Both AI-based tools achieved higher AUCs for FIV thresholds (30/50&#xa0;ml). But DeLong tests showed no statistically significant differences among packages (p &gt; 0.05 for all pairwise comparisons). Predictors of HT included age, rCBF, Tmax &gt; 6&#xa0;s, mismatch volume, and treatment; prognostic factors for mRS included rCBF, Tmax &gt; 6&#xa0;s, FIV, and HT. No significant differences were observed among packages in predicting HT or mRS (AUCs: HT 0.882–0.892; mRS 0.750–0.781).</p> Conclusion <p>The observed differences among software packages suggest that AI-augmented software offers more reliable core/penumbra quantification and FIV prediction. Although these advantages were not statistically significant in all comparisons, CTP parameters are crucial in predicting HT and functional outcome, and they exhibit comparable performance across all platforms.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Comparative analysis of three commercial CT perfusion software packages for acute ischemic stroke evaluation

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月19日

查看全文
  • Meng Gong,
  • Chao Huang,
  • Yihao Yao,
  • Shuoqi Zhang,
  • Guiling Zhang,
  • Shun Zhang

摘要

Objective

To explore the differences and the prognostic predicted value of three commercial CT perfusion (CTP) post-processing software packages (Packages A, B, and C) in quantifying ischemic core /penumbra and predicting prognoses in acute ischemic stroke (AIS) patients.

Methods

This is a retrospective analysis of 75 AIS patients processed with three software packages. Agreement was assessed via Bland–Altman plots and intraclass correlation coefficients (ICC). Multivariable regression models evaluated associations between CTP metrics and final infarct volume (FIV), hemorrhagic transformation (HT), and functional outcomes (modified Rankin Scale [mRS]). For model validation, we report the adjusted R2 for linear models and the area under the curve (AUC) with 95% confidence intervals for logistic models. No internal validation was performed, which is acknowledged as a limitation.

Results

AI-augmented packages A and B showed higher inter-package agreement and stronger correlation with FIV (rA = 0.658, rB = 0.675 vs. rC = 0.430) and better predictive accuracy for FIV (adjusted R2: A = 0.407, B = 0.398 vs. C = 0.199). Both AI-based tools achieved higher AUCs for FIV thresholds (30/50 ml). But DeLong tests showed no statistically significant differences among packages (p > 0.05 for all pairwise comparisons). Predictors of HT included age, rCBF, Tmax > 6 s, mismatch volume, and treatment; prognostic factors for mRS included rCBF, Tmax > 6 s, FIV, and HT. No significant differences were observed among packages in predicting HT or mRS (AUCs: HT 0.882–0.892; mRS 0.750–0.781).

Conclusion

The observed differences among software packages suggest that AI-augmented software offers more reliable core/penumbra quantification and FIV prediction. Although these advantages were not statistically significant in all comparisons, CTP parameters are crucial in predicting HT and functional outcome, and they exhibit comparable performance across all platforms.