Background: Low-dose computer tomography (LDCT) screening for individuals at high risk for developing lung cancer has led to earlier detection of suspicious nodules, rapid diagnosis, and a reduction in deaths due to lung cancer. Unfortunately, the majority of lung cancers are diagnosed at advanced stages where most patients will still die from their disease. We conducted a pilot study to evaluate whether chest CT scans performed routinely during emergency room (ER) visits for general medical conditions could be used to identify patients with incidental lung cancers. Methods: NLP software was used to review all CT scans obtained in 3 community-based hospital ERs in Nashville, TN and Dallas, TX from October, 2014 through December, 2015. All CT radiology reports containing key words such as: ‘nodule’, ‘opacity’, or ‘mass’ were identified by NLP software and collected by nurse navigators. Corresponding images were then reviewed by cancer center multi-disciplinary tumor boards (MDTB). Repeat imaging or biopsy was recommended based on National Comprehensive Cancer Network (NCCN) guidelines. Results: NLP software identified 1212 CT scans for MDTB review, which led to 64 biopsies (5% of NLP-identified cases). There were 37 cancer diagnoses (3% of NLP-identified cases) -26 (70%) lung cancers and 11 (30%) other cancers (including: head and neck, lymphoma, sarcoma, hepatic, breast, pancreatic, and gynecologic). Among the diagnosed lung cancers, 7 (27%) were early stage (TNM I or II) and were surgically treated for potential cure. Conclusions: The use of NLP software to review CT scans performed during routine ER encounters led to early diagnosis and successful treatment of patients whose cancers might otherwise have gone undetected until potentially much later. This pilot study is expanding to include multiple community hospitals across additional markets.