The parameterization of active faults in seismic hazard prediction still involves many uncertainties. The aim of this study is to apply logic-tree analysis, which has been widely used in engineering research, to seismic hazard prediction, in order to refine earthquake magnitude estimation. In this study, earthquake magnitude estimation takes into account the results of the Headquarters for Earthquake Research Promotion (HERP) and focuses on the Itoigawa-Shizuoka tectonic line (ISTL) using a logic-tree consisting of four branches: geometric segmentation, a coupling scenario, an empirical magnitude estimation equation and large earthquake scaling. Depending on whether the branch deals with epistemic or aleatory uncertainty, branch weight was subjectively allocated by the Technical Facilitation Integrator (TFI) under the direction of the Technical Integrator (TI), a panel consisting of six specialists. An important outcome of this study is the realization of how quantitatively different specialists' opinions are with respect to the uncertainties. The logic-tree, including weights for the individual TIs and the IF1, calculates the 50-year conditional probability of an event with seismic intensity of 6.5 or higher (JMA scale) for the cities of Matsumoto and Kofu. The results are compared with those of the HERP. Although there is no significant difference between the results for Matsumoto, differences in the results for Kofu differ by a factor of 2.2 to 5.0 times. The main reasons for the differences are: (1) the evaluation of earthquake occurrence probability under some coupling scenarios for the Gofukuji fault; and (2) uncertainty whether the total length of the ISTL could rupture simultaneously. Finally, in order to assess the influence of epistemic uncertainties (geometric segmentation, empirical magnitude estimation equation and large earthquake scaling), the branches are compared by calculating average hazard curves for each branch. The results show that the most important of the branches is large earthquake scaling.