Deep Learning Models Predict Microvascular Invasion in HCC
This meta-analysis (searched to Oct 16, 2025) pooled 52 studies with 19,531 hepatocellular carcinoma patients to evaluate imaging-based deep learning models predicting microvascular invasion. Pooled sensitivity was 0.80, specificity 0.82, and SROC 0.88, with contrast-enhanced CT achieving SROC 0.90 and pathological sections SROC 0.92. The authors note performance declines on external validation, urging prospective multicenter studies.
Key Points
- 1Report pooled sensitivity 0.80 and specificity 0.82 across 52 studies with 19,531 patients
- 2Show contrast-enhanced CT achieves highest noninvasive performance (SROC 0.90), guiding imaging choice
- 3Emphasize external validation weakness — performance drops (SROC 0.85), necessitating multicenter prospective studies
Scoring Rationale
Comprehensive meta-analysis across 52 studies gives strong evidence, limited by heterogeneity and scarce independent external validation.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems