Análise de técnicas de detecção de oclusão do escoamento de líquidos em bombas de infusão

Abstract

This work aims at proposing and evaluating obstruction detection techniques in the liquid flow of infusion pumps, using Artificial Intelligence based algorithms and comparing them with traditional approaches. We do so by applying different techniques on data collected from extensive experiments and show that the proposed techniques help in the identification of obstructions with a reduced rate of false positives and false negatives. In addition, they also require less time to detect obstructions. In particular, we propose the application of a novel algorithm based on the comparison of patterns in trends representing the temporal evolution of pressure inside the equipment. We show that this algorithm is able to outperform all other tested techniques in a variety of scenarios. This development is particularly important in situations of low flow of medications, where in the present commercial infusion pumps take a long time to detect obstructions.