An adversarial learning approach to generate pressure support ventilation waveforms for asynchrony detection

BACKGROUND AND OBJECTIVE

Mechanical ventilation is a life-saving treatment for critically-ill patients. During treatment, patient-ventilator asynchrony (PVA) can occur, which can lead to pulmonary damage, complications, and higher mortality. While traditional detection methods for PVAs rely on visual inspection by clinicians, in recent years, machine learning models are being developed to detect PVAs automatically. However, training these models requires large labeled datasets, which are difficult to obtain, as labeling is a labour-intensive and time-consuming task, requiring clinical expertise. Simulating the lung-ventilator interactions has been proposed to obtain large labeled datasets to train machine learning classifiers. However, the obtained data lacks the influence of different hardware, of servo-controlled algorithms, and different sources of noise. Here, we propose VentGAN, an adversarial learning approach to improve simulated data by learning the ventilator fingerprints from unlabeled clinical data.

METHODS

In VentGAN, the loss functions are designed to add characteristics of clinical waveforms to the generated results, while preserving the labels of the simulated waveforms. To validate VentGAN, we compare the performance for detection and classification of PVAs when training a previously developed machine learning algorithm with the original simulated data and with the data generated by VentGAN. Testing is performed on independent clinical data labeled by experts. The McNemar test is applied to evaluate statistical differences in the obtained classification accuracy.

RESULTS

VentGAN significantly improves the classification accuracy for late cycling, early cycling and normal breaths (p< 0.01); no significant difference in accuracy was observed for delayed inspirations (p = 0.2), while the accuracy decreased for ineffective efforts (p< 0.01).

CONCLUSIONS

Generation of realistic synthetic data with labels by the proposed framework is feasible and represents a promising avenue for improving training of machine learning models.

Reverse triggering during mechanical ventilation: Diagnosis and clinical implications

This review addresses the phenomenon of "reverse triggering", an asynchrony that occurs in deeply sedated patients or patients in transition from deep to light sedation. Reverse triggering has been reported to occur in 30-90% of all ventilated patients. The underlying pathophysiological mechanisms remain unclear, but "entrainment" is proposed as one of them. Detecting this asynchrony is crucial, and methods such as visual inspection, esophageal pressure, diaphragmatic ultrasound and automated methods have been used. Reverse triggering may have effects on lung and diaphragm function, probably mediated by the level of breathing effort and eccentric activation of the diaphragm. The optimal management of reverse triggering has not been established, but may include the adjustment of ventilatory parameters as well as of sedation level, and in extreme cases, neuromuscular block. It is important to understand the significance of this condition and its detection, but also to conduct dedicated research to improve its clinical management and potential effects in critically ill patients.

Bygone Ether: Theriac to Obstinate Hiccups-Food for Thought!

How to cite this article: Hirolli D, Panda R, Baidya DK. Bygone Ether: Theriac to Obstinate Hiccups-Food for Thought! Indian J Crit Care Med 2022;26(7):884.

A novel technique for assessment of post-extubation airway obstruction can successfully replace the conventional cuff leak test: a pilot study

Background

Post-extubation airway obstruction is an important complication of tracheal intubation. The cuff leak test is traditionally used to estimate the risk of this complication. However, the cuff leak test parameters are not constant and may depend on the respiratory system and ventilator settings. Furthermore, deflating the cuff also be a risk factor for patient-ventilator asynchrony and ventilator-associated pneumonia.

Instead of using the cuff leak test, we measured the pressure of the leak to the upper airway through the gap between the tube and glottis with a constant low flow from the lumen above the cuff without deflating the cuff and called it "pressure above the cuff."

The purpose of this study was to investigate whether pressure above the cuff can be used as an alternative to the cuff leak volume.

Methods

This prospective observational study was conducted at Kumamoto University Hospital after obtaining approval from the institutional review board. The pressure above the cuff was measured using an endotracheal tube with an evacuation lumen above the cuff and an automated cuff pressure modulation device. We pumped 0.16 L per minute of air and measured the steady-state pressure using an automated cuff pressure modulation device. Then, the cuff leak test was performed, and the cuff leak volume was recorded. The cuff leak volume was defined as the difference between the expiratory tidal volume with the cuff inflated and deflated. The relationship between the pressure above the cuff and cuff leak volume was evaluated. The patient-ventilator asynchrony during each measurement was also examined.

Results

The pressure above the cuff was measured, and the cuff leak volume was assessed 27 times. The pressure above the cuff was significantly correlated with the cuff leak volume (r = -0.76, p < 0.001). Patient-ventilator asynchrony was detected in 37% of measurements during the cuff leak test, but not during the pressure above the cuff test.

Conclusions

This study suggests that pressure above the cuff measurement may be a less complicated alternative to the conventional cuff leak test for evaluation of the risk of post-extubation airway obstruction.

Trial registration

University Hospital Medical Information Network Clinical Trials Registry (UMIN000039987; March 30, 2020). https://upload.umin.ac.jp/cgi-open-bin/ctr_e/ctr_view.cgi?recptno=R000044604

An interpretable 1D convolutional neural network for detecting patient-ventilator asynchrony in mechanical ventilation

BACKGROUND AND OBJECTIVE

Patient-ventilator asynchrony (PVA) is the result of a mismatch between the need of patients and the assistance provided by the ventilator during mechanical ventilation. Because the poor interaction between the patient and the ventilator is associated with inferior clinical outcomes, effort should be made to identify and correct their occurrence. Deep learning has shown promising ability in PVA detection; however, lack of network interpretability hampers its application in clinic.

METHODS

We proposed an interpretable one-dimensional convolutional neural network (1DCNN) to detect four most manifestation types of PVA (double triggering, ineffective efforts during expiration, premature cycling and delayed cycling) under pressure control ventilation mode and pressure support ventilation mode. A global average pooling (GAP) layer was incorporated with the 1DCNN model to highlight the sections of the respiratory waveform the model focused on when making a classification. Dilation convolution and batch normalization were introduced to the 1DCNN model for compensating the reduction of performance caused by the GAP layer.

RESULTS

The proposed interpretable 1DCNN exhibited comparable performance with the state-of-the-art deep learning model in PVA detection. The F1 scores for the detection of four types of PVA under pressure control ventilation and pressure support ventilation modes were greater than 0.96. The critical sections of the waveform used to detect PVA were highlighted, and found to be well consistent with the understanding of the respective type of PVA by experts.

CONCLUSIONS

The findings suggest that the proposed 1DCNN can help detect PVA, and enhance the interpretability of the classification process to help clinicians better understand the results obtained from deep learning technology.

Clinical decision support recommending ventilator settings during noninvasive ventilation

NIV therapy is used to provide positive pressure ventilation for patients. There are protocols describing what ventilator settings to use to initialize NIV; however, the guidelines for titrating ventilator settings are less specific. We developed an advisory system to recommend NIV ventilator setting titration and recorded respiratory therapist agreement rates at the bedside. We developed an algorithm (NIV advisor) to recommend when to change the non-invasive ventilator settings of IPAP, EPAP, and FiO₂ based on patient respiratory parameters. The algorithm utilized a multi-target approach; oxygenation, ventilation, and patient effort. The NIV advisor recommended ventilator settings to move the patient's respiratory parameters in a preferred target range. We implemented a pilot study evaluating the usability of the NIV advisor on 10 patients receiving critical care with non-invasive ventilation (NIV). Respiratory therapists were asked their agreement on recommendations from the NIV advisor at the patient's bedside. Bedside respiratory therapists agreed with 91% of the ventilator setting recommendations from the NIV advisor. The POB and VT values were the respiratory parameters that were most often out of the preferred target range. The IPAP ventilator setting was the setting most often considered in need of changing by the NIV advisor. The respiratory therapists agreed with the majority of the recommendations from the NIV advisor. We consider the IPAP recommendations informative in providing the respiratory therapist assistance in targeting preferred POB and Vt values, as these values were frequently out of the target ranges. This pilot implementation was unable to produce the results required to determine the value of the EPAP recommendations. The FiO₂ recommendations from the NIV advisor were treated as ancillary information behind the IPAP recommendations.

Combined effects of leaks, respiratory system properties and upper airway patency on the performance of home ventilators: a bench study

Background

Combined effects of leaks, mechanical property of respiratory system and upper airway (UA) patency on patient-ventilator synchrony (PVA) and the level of clinically “tolerable” leaks are not well established in home ventilators.

Methods

We comparatively assessed on a bench model, the highest leak level tolerated without inducing significant asynchrony (“critical leak”) in three home ventilators (Astral 150, Trilogy 100 and Vivo 60; noted as A150, T100 and V60 respectively) subjected to three simulated diseased respiratory conditions: chronic obstructive pulmonary disease (COPD), obesity hypoventilation (OHS) and neuromuscular disorders (NMD), with both open and closed UA. Also, total leak values in the device reports were compared to the bench-measured values.

Results

With open UA, all ventilators were able to avoid asynchrony up to a 30 L/min leak and even to 55 L/min in some cases. UA closure and respiratory diseases especially OHS influenced PVA. With closed UA, the critical leak of A150 and T100 remained higher than 55 L/min in COPD and OHS, while for V60 decreased to 41 and 33 L/min respectively. In NMD with closed UA, only T100 reached a high critical leak of 69 L/min. Besides, inspiratory trigger sensitivity change was often necessary to avoid PVA.

Conclusions

Home ventilators were able to avoid PVA in high-level leak conditions. However, asynchrony appeared in cases of abnormal mechanical properties of respiratory system or closed UA. In case of closed UA, the EPAP should be adjusted prior to the inspiratory trigger.

Trial registration

Not applicable.

Electronic supplementary material

The online version of this article (10.1186/s12890-017-0487-2) contains supplementary material, which is available to authorized users.

New setting of neurally adjusted ventilatory assist for noninvasive ventilation by facial mask: a physiologic study

Background

Noninvasive ventilation (NIV) is generally delivered using pneumatically-triggered and cycled-off pressure support (PS_P) through a mask. Neurally adjusted ventilatory assist (NAVA) is the only ventilatory mode that uses a non-pneumatic signal, i.e., diaphragm electrical activity (EAdi), to trigger and drive ventilator assistance. A specific setting to generate neurally controlled pressure support (PS_N) was recently proposed for delivering NIV by helmet. We compared PS_N with PS_P and NAVA during NIV using a facial mask, with respect to patient comfort, gas exchange, and patient-ventilator interaction and synchrony.

Methods

Three 30-minute trials of NIV were randomly delivered to 14 patients immediately after extubation to prevent post-extubation respiratory failure: (1) PS_P, with an inspiratory support ≥8 cmH₂O; (2) NAVA, adjusting the NAVA level to achieve a comparable peak EAdi (EAdi_peak) as during PS_P; and (3) PS_N, setting the NAVA level at 15 cmH₂O/μV with an upper airway pressure (Paw) limit to obtain the same overall Paw applied during PS_P. We assessed patient comfort, peak inspiratory flow (PIF), time to reach PIF (PIF_time), EAdi_peak, arterial blood gases, pressure-time product of the first 300 ms (PTP_300-index) and 500 ms (PTP_500-index) after initiation of patient effort, inspiratory trigger delay (Delay_TR-insp), and rate of asynchrony, determined as asynchrony index (AI%). The categorical variables were compared using the McNemar test, and continuous variables by the Friedman test followed by the Wilcoxon test with Bonferroni correction for multiple comparisons (p < 0.017).

Results

PS_N significantly improved patient comfort, compared to both PS_P (p = 0.001) and NAVA (p = 0.002), without differences between the two latter (p = 0.08). PIF (p = 0.109), EAdi_peak (p = 0.931) and gas exchange were similar between modes. Compared to PS_P and NAVA, PS_N reduced PIF_time (p < 0.001), and increased PTP_300-index (p = 0.004) and PTP_500-index (p = 0.001). NAVA and PS_N significantly reduced Delay_TR-insp, as opposed to PS_P (p < 0.001). During both NAVA and PS_N, AI% was <10% in all patients, while AI% was ≥10% in 7 patients (50%) with PS_P (p = 0.023 compared with both NAVA and PS_N).

Conclusions

Compared to both PS_P and NAVA, PS_N improved comfort and patient-ventilator interaction during NIV by facial mask. PS_N also improved synchrony, as opposed to PS_P only.

Trial registration

ClinicalTrials.gov, NCT03041402. Registered (retrospectively) on 2 February 2017.