The traditional quantitative methods for measurement of /spl alpha/-fetoprotein (AFP) are radioimmunoassay (KIA) and nonisotopic immunoassay. These immunoassay methods have the advantages of high sensitivity and accuracy. The time-consuming and complicated laboratory operations are disadvantages of these methods. On the contrary, the rapid gold immuno-chromatographic dipstick assay has the advantages of simple operation procedures and rapid operation time. But most of the dipstick immunoassays is qualitative or semi-quantitative. This paper presents a new quantitative assessment system of serum AFP by combining the rapid gold immuno-chromatographic dipstick assay and neural network based computer image analysis system. A total of 72 human serum samples, in which the AFP concentrations were measured by RIA as 1 to 700 ng/ml, are used to construct the neural network for quantitative computer image analysis system in this study. In this paper, we also correlated the proposed system results with a commercial AFP specific RIA test kit. Another 30 human serum samples were used to validate the new system. The result of the new system correlated well with traditional RIA with correlation coefficient 0.9971. The Bland-Altman plot also shows well agreements between the proposed system and RlA. We successfully established a rapid quantitative assessment system for AFP dipstick assay, which took significantly less time in operation, with clinical acceptable agreements by Bland-Altman plot analysis. This novel method is simple, rapid, and can be performed outside the laboratory. The rapid gold immuno-chromatographic dipstick assay quantified by computer image analysis has been termed dipstick computer image processing immunoassay (DClPA).