Deep Learning Model With Nodule Indexing Tailored to Early-Stage Lung Cancer Detection.

Abstract

The aim of this study was to evaluate whether a deep learning-based artificial intelligence (AI) system with suspected nodule indexing and malignancy risk stratification improves radiologist performance in detecting pulmonary nodules on CT, using a dataset enriched with challenging early-stage lung cancers. The study comprised a stand-alone AI sensitivity-specificity analysis and a two-arm crossover reader study with 16 American board-certified radiologists. Each reader interpreted 340 CT scans with and without AI, separated by a 1-month washout period. The dataset included 209 screening and 131 nonscreening cases: 133 with lung cancer, 61 with benign noncalcified nodules ≥4 mm, and 146 normal. To enrich subtle lesions, 64 of 91 small cancer cases (70.3%) were drawn from early-round National Lung Screening Trial (NLST) CT scans. Localization-specific receiver operating characteristic analysis was used to assess radiologist performance. Stand-alone AI achieved sensitivity of 0.804 at 1.37 false positives per case. With AI assistance, radiologists' area under the localization-specific receiver operating characteristic curve (AUC) improved cancer detection (0.761 vs 0.652; ΔAUC = 0.109; 95% confidence interval [CI]: 0.067 to 0.152) and for all nodules (0.830 vs 0.734; ΔAUC = 0.096; 95% CI: 0.059 to 0.133). Mean sensitivity increased from 0.585 to 0.727, while specificity remained essentially unchanged (0.918 vs 0.913). Interpretation time decreased by 12.9%, from a mean of 133 to 115.9 seconds (difference = -17.1 seconds; 95% CI: -26.7 to -9.0 seconds). AI alerts enabled the detection of early-stage cancer detection previously missed in NLST interpretations. The AI system significantly improved radiologists' performance in pulmonary nodule detection, with consistent benefits across nodule types, screening contexts, and experience levels, supporting its integration into routine chest CT interpretation workflows.