Total Nodule Number As an Independent Prognostic Factor in Resected Stage III Non-Small Cell Lung Cancer: a Deep Learning-Powered Study

Almost every lung cancer patient has multiple pulmonary nodules while the significance of nodule multiplicity in locally advanced non-small cell lung cancer (NSCLC) remained unclear. This study explores the relationship between deep learning detected total nodule number (TNN) and survival outcomes in patients with surgical resected stage I-III NSCLC. Patients who underwent surgical resection for stage I-III NSCLC with accessible preoperative chest CT scan from 2005 to 2018 were identified from our database. Deep learning-based AI algorithms using convolutional neural networks (CNN) was applied for pulmonary nodule (PN) detection and classification. Of the 2126 patients, a total number of 33410 PNs were detected by AI. Median TNN detected per person was 12 (IQR 7-20). AI-detected TNN (analyzed as continuous variable) was independent prognostic factor for both RFS (HR 1.012, 95% CI 1.002-1.022, p = 0.021) and OS (HR 1.013, 95% CI 1.002-1.025, p = 0.021) in multivariate analyses of stage III cohort; while it was not significantly associated with survival in stage I and II cohorts. In terms of nodule categories, the numbers of upper-lobe nodule, same-side nodule, other-side nodule, solid nodule, and even solid nodule at small size ([≤]6mm) were independent prognostic factors; while the numbers of middle/lower-lobe nodule, same-lobe nodule, subsolid nodule, calcific nodule and perifissural nodule were not associated with survival. In survival tree analysis, rather than using traditional IIIA and IIIB classification, the model grouped cases by AI-detected TNN (lower vs. higher: log-rank p < 0.001), which showed superior discrimination of survival in stage III cohort. In conclusion, AI-detected TNN was significantly associated with survival in patients with surgical resected stage III NSCLC. Lower TNN detected on preoperative CT scan indicated better prognosis in patients who underwent complete surgical resection.