In this paper, we consider a decode-and-forward (DF) relay-assisted diffusion-based molecular communication system inside one of the blood vessels of a human body with positive drift from transmitter to receiver. We use the normal approximation to the distribution of the number of received molecules and derive a closed-form expression for the end-to-end bit error probability of the system. We then propose an optimization problem that aims at minimizing the bit error probability of the system and solve it at the receiver nanomachine by an algorithm based on the bisection method to determine the optimal detection threshold. Furthermore, we study the impact of the system parameters, such as drift velocity, position of the relay node and number of allocated molecules on the performance of the system. The numerical results show that with a constant molecular budget, DF relying strategy can considerably improve the system performance.