Preoperative prediction of hepatocellular carcinoma histological grade using MRI-based artificial intelligence models: A systematic review and meta-analysis.
Longshan Yang, Zeyi Zhang, Yanchen Dong et al.
✦ AI-curated · Sources linked
In 30 seconds
This systematic review and meta-analysis evaluated the diagnostic performance of MRI-based artificial intelligence models for predicting histological grade in hepatocellular carcinoma (HCC). The analysis included 18 studies, revealing that in internal validation, the pooled sensitivity was 0.78 and specificity was 0.80 for high-grade HCC prediction. However, external validation showed lower sensitivity and specificity at 0.70 and 0.74, respectively.
Key findings
- In internal validation, pooled sensitivity was 0.78 and specificity was 0.80 for predicting high-grade HCC.
- External validation showed pooled sensitivity of 0.70 and specificity of 0.74.
- Deep learning models had significantly higher sensitivity than machine learning models (0.88 vs. 0.72, P = 0.018).
- At a 20% pre-test probability, positive post-test probabilities were 48% (internal) and 40% (external).
Why it matters
The ability to accurately predict the histological grade of HCC preoperatively can guide treatment decisions and improve patient outcomes. However, the current evidence suggests caution in clinical application due to low certainty and heterogeneity.
What to keep in mind
The study highlights very low evidence certainty and significant heterogeneity among the included studies.