RESUMO
BACKGROUNDS: Substituting nebulisers by another, especially in non-invasive ventilation (NIV), involves many process-variables, e.g. nebulizer-type and fill-volume of respirable-dose, which might affect patient optimum-therapy. The aim of the present work was to use neural-networks and genetic-algorithms to develop performance-models for two different nebulizers. METHODS: In-vitro, ex-vivo and in-vivo models were developed using input-variables including nebulizer-type [jet nebulizer (JN) and vibrating mesh nebulizer (VMN)] fill-volumes of respirable dose placed in the nebulization chamber with an output-variable e.g. average amount reaching NIV patient. Produced models were tested and validated to ensure effective predictivity and validity in further optimization of nebulization process. RESULTS: Data-mining produced models showed excellent training, testing and validation correlation-coefficients. VMN showed high nebulization efficacy than JN. JN was affected more by increasing the fill-volume. The optimization process and contour-lines obtained for in-vivo model showed increase in pulmonary-bioavailability and systemic-absorption with VMN and 2â¯mL fill-volumes. CONCLUSIONS: Modeling of aerosol-delivery by JN and VMN using different fill-volumes in NIV circuit was successful in demonstrating the effect of different variable on dose-delivery to NIV patient. Artificial neural networks model showed that VMN increased pulmonary-bioavailability and systemic-absorption compared to JN. VMN was less affected by fill-volume change compared to JN which should be diluted to increase delivery.
