This study aims to implement deep neural networks (DNNs) to assess the sensitivity level of design parameters in flexible pavements for roads with low traffic volume. One hundred eight structures were modeled using the Pitra Pave® software (i.e., multilayer elastic theory (MET) model for pavement structural analysis) to generate a database to develop the DNN models. The DNN models, through connection weights, allowed the comparison with MET to evaluate the sensitivity of the selected design variables (resilient modulus and layer thickness) on the structural design parameters. The results suggest the significant impact of layer thicknesses. In addition, the predictions of structural design parameters from these initial DNN models showed variations ranging from 0,03% to 10,87% compared to MET. Expanding the database and developing a multi-predictive network is recommended for future research.