We have developed a computer method which by automatically locating the border of the gallbladder in each image and subtracting a varying location-dependent background corresponding to the gallbladder border overcomes the difficulties associated with quantitating gallbladder contraction. These difficulties are attributable to significant and changing background activity, imprecise manual delineation of the gallbladder, and the changing position and shape of the gallbladder. Validation studies using a gallbladder phantom showed the method to be very accurate. No significant difference (P greater than 0.05) was observed between the expected and calculated ejection fractions. Ten patient studies were analyzed, with a resulting range of gallbladder ejection fractions of 22-79%. Excellent reproducibility was obtained with an average intraobserver coefficient of variation of 3.5%, and no statistically significant difference in interobserver measurements (P greater than 0.05). The regression line for interobserver measurements had a slope of 0.96 +/- 0.08, an intercept of 1.6 +/- 4.0%, and a correlation coefficient of 0.99. From these preliminary results we conclude that this method offers a reliable means of quantitating gallbladder contraction.