Background and Objective: The background for this work was several Literature reports on applications of the fluorescence methods to detection and localization of human cancers. The objective of our study has been to investigate if such an approach could be applied for the detection of gastric cancers. Study Design/Materials and Methods: Our study was designed in such a way that spectrally resolved images of laser-induced fluorescence of human gastric mucosa were collected and assessed from a point of view of elaborating an algorithm allowing for a differentiation between malignant and premalignant lesions and areas of normal mucosa. The method involved exciting the autofluorescence with ultraviolet light (325 nm, He-Cd laser). The images were recorded in vitro in six regions of a visible spectrum using a cooled CCD camera. The material for study was 21 resected specimens for which altogether 72 surface areas were examined. Results: The main result is the observation that a difference of the fluorescence intensities measured at 440 nm and 395 nm, both normalized to intensity measured at 590 nm, differs significantly for the tissues of interest. Conclusion: Using that difference as a diagnostic parameter, it was possible to classify malignant tumor tissues with a sensitivity of 96% and a predictive value of 42%, whereas the same approach applied to abnormal but not tumor stomach tissues gave values of 80% and 98%, respectively. (C) 1997 Wiley-Liss, inc.