A new estimation method of clinical blood indices based on the flow characteristics of unadjusted whole blood is proposed, such as red-cell deformability and plasma viscosity. Parameter values of Bingham and power law models were calculated based on the measured flow characteristics. Red-cell deformability was measured by the method of laser diffractometry. The clinical indices were related with the non-Newtonian model parameters by a feedforward neural network (NN). Inputs of the NN are the parameter values and hematocrit. Outputs are red-cell deformability and plasma viscosity. It was verified that estimated values agreed well with those measured values after sufficient learning. By applying this method to 20 healthy persons, it was confirmed that the estimated values by NN were within normal range. According to these experimental results, this estimation method cart be applied to clinical use.