Diagnostic accuracy of AI-Based models for pulmonary edema detection: A systematic review and Meta-Analysis.
Pooya Eini, Peyman Eini, Homa Serpoush et al.
✦ AI-curated · Sources linked
In 30 seconds
This systematic review and meta-analysis evaluated the diagnostic accuracy of machine learning models for detecting pulmonary edema across 14 studies. The models showed high pooled sensitivity of 0.90 and specificity of 0.89, indicating strong potential for early detection. However, substantial heterogeneity (I² = 79.36%) was noted, affecting generalizability.
Key findings
- Pooled sensitivity of machine learning models for pulmonary edema detection was 0.90 (95% CI: 0.78-0.96).
- Pooled specificity was 0.89 (95% CI: 0.70-0.96).
- The area under the receiver operating characteristic curve (AUROC) was 0.95 (95% CI: 0.93-0.97).
- Substantial heterogeneity (I² = 79.36%) was observed across studies.
Why it matters
Improving the early detection of pulmonary edema is crucial for patient outcomes. The promising diagnostic accuracy of AI models could enhance clinical decision-making, although variability in study results highlights the need for further research.
What to keep in mind
Substantial heterogeneity across studies limited the generalizability of findings, and the certainty of evidence was graded as low.