Scholarlink

<p>Early prediction of hematoma expansion (HE) in patients with intracerebral hemorrhage (ICH) is critical for improving clinical outcome and guiding timely interventions. This study focuses on assessing the effectiveness of artificial intelligence (AI) models, specifically those utilizing machine learning (ML) and deep learning (DL) in predicting HE. The search strategy for this study was conducted in PubMed, Scopus, Embase, and Web of Science to identify eligible studies. Metrics such as sensitivity, specificity and area under the curve (AUC) were extracted and analyzed. Out of 1,235 studies initially screened, 14 met the inclusion criteria with totally 7665 patients. AI algorithms demonstrated significant predictive capabilities. The pooled sensitivity and specificity across studies were 0.82 [95% CI: 0.74–0.88] and 0.83 [95% CI: 0.78–0.87], respectively. The pooled positive DLR was 4.91 [95% CI: 3.73–6.47], while the negative DLR was 0.21 [95% CI: 0.14–0.32]. The diagnostic score was 3.14 [95% CI: 2.55–3.72], and the overall diagnostic odds ratio was 23.01 [95% CI: 12.83–41.24]. The pooled AUC was 0.89, suggesting a potentially useful diagnostic performance, although the high heterogeneity limits the robustness of this finding. Subgroup analysis revealed that AI models performed better in predicting HE, in patients with spontaneous ICH compared to patients with traumatic brain injury-related ICH, with a higher AUC (0.90 vs. 0.87), sensitivity (0.85 vs. 0.76) and diagnostic odds ratio (28 vs. 16). AI-based models may serve as potentially useful tools in predicting HE and supporting clinical decisions in ICH care. However, due to the high heterogeneity across studies, these findings should be interpreted with caution, and rigorous external validation with standardized imaging protocols is essential before clinical implementation.</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

Can artificial intelligence accurately predict the risk of hematoma expansion in intracerebral hemorrhage? A systematic review and Meta-analysis of 7,665 patients

  • 期刊论文

  • 发表日期：

    2025年11月24日

查看全文
  • Ibrahim Mohammadzadeh,
  • Bardia Hajikarimloo,
  • Shahin Mohammadzadeh,
  • Amin Mohamad Niaei,
  • Paniz Sanjari Pirayvatloo,
  • Mohammad Amin Habibi,
  • Poriya Minaee,
  • Adam A. Dmytriw,
  • Ahmet Günkan,
  • Pascal Jabbour

摘要

Early prediction of hematoma expansion (HE) in patients with intracerebral hemorrhage (ICH) is critical for improving clinical outcome and guiding timely interventions. This study focuses on assessing the effectiveness of artificial intelligence (AI) models, specifically those utilizing machine learning (ML) and deep learning (DL) in predicting HE. The search strategy for this study was conducted in PubMed, Scopus, Embase, and Web of Science to identify eligible studies. Metrics such as sensitivity, specificity and area under the curve (AUC) were extracted and analyzed. Out of 1,235 studies initially screened, 14 met the inclusion criteria with totally 7665 patients. AI algorithms demonstrated significant predictive capabilities. The pooled sensitivity and specificity across studies were 0.82 [95% CI: 0.74–0.88] and 0.83 [95% CI: 0.78–0.87], respectively. The pooled positive DLR was 4.91 [95% CI: 3.73–6.47], while the negative DLR was 0.21 [95% CI: 0.14–0.32]. The diagnostic score was 3.14 [95% CI: 2.55–3.72], and the overall diagnostic odds ratio was 23.01 [95% CI: 12.83–41.24]. The pooled AUC was 0.89, suggesting a potentially useful diagnostic performance, although the high heterogeneity limits the robustness of this finding. Subgroup analysis revealed that AI models performed better in predicting HE, in patients with spontaneous ICH compared to patients with traumatic brain injury-related ICH, with a higher AUC (0.90 vs. 0.87), sensitivity (0.85 vs. 0.76) and diagnostic odds ratio (28 vs. 16). AI-based models may serve as potentially useful tools in predicting HE and supporting clinical decisions in ICH care. However, due to the high heterogeneity across studies, these findings should be interpreted with caution, and rigorous external validation with standardized imaging protocols is essential before clinical implementation.