A total of 152 normal bronchial biopsy sections--63 from normal subjects, 42 from patients with dysplasia, 28 from patients with early or advanced lung cancer with squamous and nonsquamous histopathology, and 19 from resected lung cancer patients--were examined for the presence of malignancy-associated changes (MACs). A standard, white light bronchoscope examination and a multispectral fluorescence bronchoscope examination were performed on every subject to determine the status of the bronchial epithelial tissue. Any suspect areas were biopsied to determine the status of the area and to establish the highest grade of abnormality in the patient. In addition, for every subject, a bronchoscopically normal area of the bronchus in the opposite lung or another lobe was biopsied. The specimens were confirmed to be normal by conventional histopathologic criteria using hematoxylin and eosin stain. Sections of biopsies were stained using a DNA stoichiometric stain, and approximately 250 images of visually normal epithelial cell nuclei from each of the biopsies were collected, as were approximately 40 images of leukocyte cell nuclei. For each of these images > 60 nuclear features were calculated that quantified the size, shape and DNA volume of the nuclei as well as DNA spatial organization in the nuclei. The features were then used to train an automated classifier to recognize normal epithelial cell nuclei from normal subjects and normal-appearing epithelial cell nuclei (MAC cell nuclei) from lung cancer patients.(ABSTRACT TRUNCATED AT 250 WORDS)