Estimation of red cell deformability and plasma viscosity by a neural network is examined for clinical application, which is based on a flow curve in the high shear rate range of 1,000 - 10,000 s(-1) obtained by a blood viscometer presently employed in a hospital. Newly defined indices DI15 and DI40 as quantified deformability were proposed in this study, which reflect the elongation of red cells under shear stress of 15 Pa and 40 Pa, respectively. It is found that a combination of non-Newtonian model parameter values gives a unique set of those quantified deformability indices and plasma viscosity under the assumption that the volume fraction of red cells (Hct) is known. A feed forward neural network newly constructed in this study, has inputs of Hct, Bingham viscosity, yield stress, exponent of power law model and outputs of those deformability indices and plasma viscosity. Though practical accuracy was not attained for the estimation of plasma viscosity, it is expected that clinical application of this method is possible for the estimation of DI15 and DI40.