Table 1 Comparison of some automated methods for patient-ventilator asynchrony detection
| Â | Type of PVA | Algorithm | Performance |
|---|---|---|---|
Gholami et al. (2018) [69] | Cycling asynchrony (premature and delayed cycling) | ML: Random forest and k-fold cross validation Pressure and airflow signals N = 11 patients (1377 breaths) | Se 89–97%, Sp 93–99%, Kappa index 0.9 |
ventMAP platform Adams et al. (2017) [70] | Double-trigger and breath stacking | Rule-based algorithm Pressure and airflow signals Derivation cohort, N = 16 patients (5075 breaths); validation cohort, N = 17 patients (4644 breaths) | Se 94–96.7%, Sp 92–98%, Acc 92.2–97.7% (on the validation cohort) |
NeuroSync index Sinderby et al. (2013) [71] | Patient-ventilator interaction classification (asynchronous, dyssynchronous or synchronous) | Rule-based timings algorithm EAdi and pressure signals N = 24 patients | ICC 0.95 vs. Colombo et al. (2011) [5] |
Better Care® system Blanch et al. (2012) [37] | Ineffective efforts during expiration | Rule-based combining digital signal processing techniques and ROC curves Airflow signal Cohort 1: N = 8 patients (1024 breaths) Cohort 2: N = 8 patients (9600 breaths) with EAdi signal as reference | Se 91.5%, Sp 91.7%, PPV 80.3%, NPV 96.7%, Kappa index 0.797 (vs. the expert’s classification) Se 65.2%, Sp 99.3%, PPV 90.8%, NPV 96.5%, Kappa index 0.739 (vs. EAdi signal) |
Gutierrez et al. (2011) [72] | Index for asynchronous/no asynchronous breaths | Time-frequency analysis Airflow signals N = 110 patients | Se 83%, Sp 83% when index < 43% for AI > 10% |
Mulqueeny et al. (2007) [73] | Ineffective triggering and double triggering | Rule-based and digital signal processing methods Airflow and pressure signals N = 20 patients (3343 breaths) | Se 91%, Sp 97% |
PVI monitor Younes et al. (2007) [74] | Ineffective efforts | Rule-based Equation of motion from pressure, airflow, and Peso signals N = 21 patients | Se 79.7% |