Learning neural impulsive MPC for tailoring therapies in viral infections

Abstract

Viruses are among the most common causes of human infections causing a high mortality rate. Thus, tailoring therapies against viral infections are central. In view of that, this article presents the scheduling of antiviral treatments for influenza A virus (IAV) infections based on neural networks and model predictive control (MPC). This hybrid intelligent approach calculates the current drug administration after using the differential evolution (DE) algorithm to establish a set of drug-intake times which maximizes the viral clearance while minimizing the drug consumption during a prediction horizon. Monte Carlo simulations reveal that the viral elimination achieved by the proposed strategy is similar to that obtained by the clinical recommendations. However, the amount of drug administrated by the algorithm is less than the current clinical recommendations. Thus, the potential of the proposed algorithm to schedule treatments in influenza and other infections is discussed.