Super-resolution CT reconstruction boosts AI lung nodule detection in thick-slice scans

Super-resolution reconstruction of 5-mm chest CT lifted artificial intelligence lung nodule sensitivity from 31.4% to 61.0% and positive predictive value from 42.9% to 80.0% in 96 colorectal cancer patients (20% hallucinated nodules).

• Retrospective multicenter study in 96 colorectal cancer patients with 105 reference pulmonary nodules <10 mm, comparing original thin-section CT, 5-mm thick CT, and super-resolution converted images using a commercial AI-CAD system.
• Nodule-level AI sensitivity improved from 31.4% with thick-section images to 61.0% with super-resolution images; PPV improved from 42.9% to 80.0%; patient-level sensitivity improved from 41.5% to 67.1%.
• Super-resolution preserved nodule morphology (95.3% anatomic concordance, 80% solid-feature consistency) and reduced image noise (P<0.001), but generated 20% hallucinated nodules (16/80), predominantly benign or artifactual.

Automated summary

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