Photo Credit: Master_Video
The following is a summary of “Automated characterization of patient-ventilator interaction using surface electromyography,” published in the February 2024 issue of Critical Care by Sauer et al.
Researchers conducted a retrospective study to Streamline the characterization of patient-ventilator interaction in critically ill patients by developing an efficient, automated system that eliminates the need for extensive time and trained personnel.
They conducted a study wherein surface electromyography (sEMG) signals from the diaphragm and intercostal muscles, along with esophageal pressure (Pes), were recorded in mechanically ventilated patients with ARDS. The sEMG recordings underwent preprocessing, followed by applying two distinct algorithms (the triangle and adaptive thresholding algorithms) to detect inspiratory patient effort automatically. Major asynchronies were computationally classified based on the detected inspirations, including ineffective, auto-, and double triggers, double efforts, and delayed and synchronous triggers. Reverse triggers were not taken into consideration in this study. Asynchrony indices were subsequently computed. To validate the detected efforts, two experts independently annotated inspiratory patient activity in Pes, sEMG signals, and the algorithmic results, with each expert blinded to the other. Additionally, patient-ventilator interaction was classified, and asynchrony indices were calculated using manually detected inspirations in Pes as a reference for automated asynchrony classification and asynchrony index calculation.
The results showed spontaneous breathing activity was observed in 22/36 patients analyzed in the study. Assessment of the algorithms’ accuracy using 3057 inspiratory efforts in Pes demonstrated consistent detection performance for both methods. Across all datasets, a notable sensitivity was observed (triangle algorithm/adaptive thresholding algorithm: 0.93/0.97), along with a high positive predictive value (0.94/0.89) compared to expert annotations in Pes. The average delay from automatically detected inspiratory onset to the Pes reference was consistent at —79 ms/29 ms for the two algorithms. Moreover, automatic prediction of the asynchrony index was found to be reliable, with both algorithms showing a consistent deviation of 0.06 ± 0.13 from the Pes-based reference.
Investigators concluded that automating asynchrony quantification with noninvasive sEMG holds promise for frequent diagnosis and improved patient-ventilator interaction.
Source: annalsofintensivecare.springeropen.com/articles/10.1186/s13613-024-01259-5
