Can incorrect artificial intelligence (AI) results impact radiologists, and if so, what can we do about it? A multi-reader pilot study of lung cancer detection with chest radiography.

Eur Radiol · 2023 · PMC10235827 · PMID 37266657

Europe PMC DOI

hedged sentences

0.0520

closest p · 1.0× alpha

0.1900

boldest claim

The sentences

did not quite achieve significancep = 0.052so close (0.05 < p ≤ 0.1)

Among these cases in the No AI condition, the false positive percent was 51.4% (95% CI (19.4, 82.3)); false positives increased to 86.0% (95% CI (67.5, 94.8)) in the AI Keep (No Box) condition ( p = 0.01), increased to 80.5% (95% CI (59.0, 92.2)) in the AI Delete (No Box) condition ( p = 0.008), and increased to 80.5% (95% CI (52.7, 93.8)) in the AI Keep (Box) condition, though this did not quite achieve significance ( p = 0.052).

also in 12 other papers

failed to achieve significancep = 0.13not close (p > 0.1)

26.7%), though this difference failed to achieve significance ( p = 0.13).

also in 179 other papers

did not achieve significancep = 0.19not close (p > 0.1)

80.5%), though this did not achieve significance ( p = 0.19), and it had an identical value to the Delete (No Box) condition (i.e., both 80.5%).

also in 816 other papers

Quoted from the open-access full text in Europe PMC under the licence the publisher applied. The sentence is reproduced exactly as published; the emphasis is ours.