Diagnostic performance of artificial intelligence based on PET/CT for lymph node involvement in prostate cancer: A systematic review and meta-analysis.
Dongxuan Song, Zhiyu Zhao, Kuangqi Chen
✦ AI-curated · Sources linked
In 30 seconds
This systematic review and meta-analysis evaluated the diagnostic performance of PET/CT-based artificial intelligence (AI) models for detecting lymph node involvement in prostate cancer, comparing them to physician readers. The analysis included 11 studies with 1,143 patients, revealing that AI had a sensitivity of 0.71 and specificity of 0.89, while physician readers had a sensitivity of 0.67 and specificity of 0.88. AI demonstrated a higher AUC of 0.79 compared to 0.71 for physicians.
Key findings
- AI sensitivity was 0.71 (95% CI: 0.61-0.79) and specificity was 0.89 (95% CI: 0.76-0.95).
- Physician readers had a sensitivity of 0.67 (95% CI: 0.57-0.76) and specificity of 0.88 (95% CI: 0.70-0.95).
- AI achieved a higher AUC of 0.79 (95% CI: 0.75-0.82) compared to 0.71 (95% CI: 0.67-0.75) for physicians.
- The study included 11 studies with a total of 1,143 patients.
Why it matters
The findings suggest that while AI models for PET/CT may improve diagnostic accuracy for lymph node involvement in prostate cancer, the current evidence does not confirm their clinical superiority over experienced physician readers. This highlights the need for further research before integrating AI into routine clinical practice.
What to keep in mind
The evidence is limited by small study numbers, mostly retrospective designs, and low certainty.