PURPOSE: To analyze the causes of delayed detection of lung cancer on chest radiographs. MATERIALS AND METHODS: We retrospectively reviewed 105 cases in which an initial diagnosis of lung cancer, based on anexamination of plain radiographs, had been missed or misinterpreted. All occurred between October 1993 and April1997. We reviewed the initial chest radiographs and compared the features noted with those seen on later chestradiographs and computed tomographic (CT) images. RESULTS: Undetected lung cancer was identified in 56 patients(56/105, 53.3%) It had been hidden by superim-posed structures (41, 73.2%), overlapped by combined benign diseases(12, 21.4%), or the nodules were subtle(3, 5.4%). Of the 41 lung cancers hidden by a superimposed structure, thecentral type accounted for 29 (70.7%) and the peripheral type for 12 (29.3%). The 29 central type had been hiddenby the left hilum (n=15), the right hilum (n=10), the heart (n=3), or a rib (n=1). The twelve peripheral type werehidden by a rib (n=7), the heart (n=2), the diaphragm (n=2), or the left hilum (n=1). Of the 12 lung cancersoverlapped by combined benign diseases, pulmonary tuberculosis (n=6), pleural effusion (n=4), congestive heartfailure (n=1), and dif-fuse interstitial lung disease (n=1) were present at the time of interpretation. Themisinterpreted lung cancers were identified in 49 patients (49/105, 46.7%) and were seen to be combined withbenign disease (16, 32.6%), or as obstructive pneumonia without a central mass (15, 30.6%), air-spaceconsolidation (7, 14.3%), cavity (7, 14.3%), double lesion (2, 4.1%), or young age below 26 years (2, 4.1%). Ofthe 16 lung cancers misinterpreted as combined disease, pulmonary tuberculosis (n=14) and pleural disease (n=2)had been initially diagnosed. CONCLUSION: Most commonly, lung cancer was missed or misinterpreted because it washidden by a normal structure or combined with a benign disease. Perceptual errors can be reduced by appropriatetechniques and the scrutiny of trouble spots such as the parahilar, retrocardiac, retrodiaphragmatic and costalregions. Errors in the analysis of lung cancer can be reduced by increased awareness of growth patterns anduncommon man-ifestations of the disease.