The film digitizer plays an important transitory role as an analog-to-digital bridge for the implementation of PACS on screen-film mammography. The objective of this phantom study was to compare the observer performance of a digitized mammogram on liquid crystal displays (LCDs) with that of the original film mammogram, and to clarify which matrix size of LCDs is adequate for the interpretation of digitized mammography. The American College of Radiology mammographic accreditation phantom containing variously sized fibrils, simulated microcalcifications, and tumor-like masses was radiographed with a screen-film mammographic system. The original film was digitized with a sampling pitch of 50 mum and a density depth of 12 bit. Six observers who were trained in mammography individually viewed digitized images on LCDs and original film. Observer performance of a digitized mammogram with a 2-megapixel LCD was compared with that of original film. The observers were asked to rate the detectability of each test object according to a three-point scale (obviously visible=2, barely detectable=1, undetectable=0). The difference in the mean score between two systems at each object was evaluated by the Wilcoxon's test. In addition, the dependence of observer performance on the matrix size of LCDs ranging from 1-to 5-megapixel was tested with Scheffé's multiple comparison. The observers also judged the detectability according to the three-point scale. The results show that the digitized mammogram provides acceptable but slightly inferior detectability than original film. There was no dependence of matrix size in observer performance with more than 2-megapixel LCDs.