K-Means Clustering of Alae Nasi and Diaphragmatic Muscles Activation Timing as an Indicator to Inspiratory Effort Level: a Proof of Concept

Abstract

The objective of this paper is to develop a non-invasive robust indicator to the inspiratory effort of a patient under mechanical ventilation. This indicator leads to the inspiratory effort detection as well as to the estimation of its level more reliably and earlier than the classical systems based on flow signal thresholding. Hence, the present work analyses the capability of inspiratory effort level estimation by the observation of the synchronization of the Alae Nasi and the parietal diaphragmatic muscles activations. First, an experimental protocol is suggested to simulate the patient-ventilator coupling. Then, the evolution of muscular activation timing -versus the inspiratory effort level- is studied on acquired ElectroMyoGraphy and flow signals. Finally, a multidimensional clustering approach is applied in order to separate timing features into classes indicating the different effort levels. The results indicate the efficacy of effort level classification in fifty-two respiratory cycles by k-means clustering based on the muscular timing features where weak effort is completely identified separately from strong effort through discrimination of centroid and distance values. The weakness/strength of inspiratory effort is judged by the energy of the parietal activity as well as by the level of expired Co2.