Comment: AI-CT Lung Nodule Review Lacks Search Strategy, Overstates Predictive Performance

A comment on a review linking AI-derived CT features of lung ground-glass nodules to driver genes notes missing search methods, selective accuracy reporting (AUC 0.64 to >0.95), and framing that overstates AI's performance given a referenced systematic review's mean AUC of ~0.64.

• The review did not report a search strategy (databases, date range, inclusion criteria) despite citing over 150 references.
• Individual studies are cited by their best accuracy/AUC (range ~0.64 to >0.95) without weighting for sample size or validation.
• The review's 'transformative' framing contrasts with a referenced systematic review's mean AUC of ~0.64 for driver gene prediction from imaging.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.