New acoustic monitoring system quantifying aspiration risk during monitored anaesthesia care

Abstract

Respiratory monitoring is crucial during monitored anaesthesia care (MAC) to ensure patient safety. Patients undergoing procedures like gastrointestinal endoscopy and dental interventions under MAC have a heightened risk of aspiration. Despite the risks, no current system or device can evaluate aspiration risk. This study presents a novel acoustic monitoring system designed to detect fluid retention in the upper airway during MAC. We conducted a prospective observational study with 60 participants undergoing dental treatment under MAC. We utilized a prototype acoustic monitoring system to assess fluid retention in the upper airway by analysing inspiratory sounds. Water was introduced intraorally in participants to simulate fluid retention; artificial intelligence (AI) analysed respiratory sounds pre and post-injection. We also compared respiratory sounds pre-treatment and during coughing events. Coughing was observed in 14 patients during MAC, and 31 instances of apnoea were detected by capnography. However, 27 of these cases had breath sounds. Notably, with intraoral water injection, the Stridor Quantitative Value (STQV) significantly increased; furthermore, the STQV was substantially higher immediately post-coughing in patients who coughed during MAC. In summary, the innovative acoustic monitoring system using AI provides accurate evaluations of fluid retention in the upper airway, offering potential to mitigate aspiration risks during MAC.Clinical trial number: jRCTs 062220054.

Figure 1

The cervical sensor is placed just lateral to the thyroid cartilage and medial to the sternocleidomastoid muscle (on the right side, indicated by a red arrow). The RRa sensor is attached contralaterally (on the left side).

Figure 2

Anaesthesia and data collection protocol diagram. STQV assessments were performed in a quiet environment to avoid noise effects. MAC monitored anaesthesia care, STQV stridor quantitative value.

Figure 3

The respiratory sound monitoring system comprises sensors for detecting respiratory sounds and the PC for visualization.

Figure 4

Machine learning algorithms for the values of the strider component. Based on a training dataset of strider sounds recorded from patients and labelled by experts, features were extracted using frequency and cepstrum analyses, and machine learning algorithms were applied.

Figure 5

Breakdown of individuals with apnoea detected by capnogram. Respiratory sounds were confirmed using audiograms and spectrograms.

Figure 6

Comparison of the values of STQV before and after intraoral injection of 3 cc of water. The STQV was significantly higher after intraoral water injection than that before (0.136 ± 0.103 vs. 0.05 ± 0.045, P < 0.001 [95% confidence interval (CI) − 0.11 to − 0.061]. The horizontal line indicates the mean, and error bar indicates the standard deviation; analysis was performed using the student’s t-test. STQV stridor quantitative value.

Figure 7

Comparison of the STQV in patients with cough. The STQV pre-treatment and immediately after the onset of coughing in patients who developed cough during MAC (n = 14) was 0.042 ± 0.04 and 0.342 ± 0.038, respectively, indicating that the STQV was significantly higher immediately after the onset of coughing (P < 0.002 [95% CI − 0.47 to − 0.13]. The horizontal line indicates the mean, and error bar indicate standard deviation; analysis was performed using the student’s t-test. STQV stridor quantitative value.