Efficient controlled release systems are fundamental for efficient therapeutical methodologies mainly to improve drug treatment through rate and time programmed drug delivery. However, the occurrence of multi components instead of single drug diffusion may complicate the ex- perimental procedure to produce adequate drug delivery systems. Therefore, mathematical models of controlled delivery systems are important to simulate the device properties such as the geometries and physicochemical properties of the system.
This work presents an alternative methodology, based on artificial neural networks to study drug delivery systems. A drug model system of hydrocortisone was selected to test the present approach but introducing modifications to increase complexity. The results of the neural network obtained after the learning stage can be considered quantitative to predict an ideal experimental condition. The analysis of the outcome results has shown that one can predict properties of an efficient experiment with a relative average error smaller than 2%.

*Work supported by CAPES, FAPEMIG and CNPQ, Brazil.