Automated detection and classification of patient–ventilator asynchrony by means of machine learning and simulated data

Bakes, Tom; Diepen, Anouk van; Bie, Ashley J.R. De; Montenij, Leon J.; Mojoli, Francesco; Bouwman, R. Arthur; Mischi, Massimo; Woerlee, P.H.; Turco, Simona

doi:10.1016/j.cmpb.2022.107333

Cited by 9 publications

(6 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Firstly, it introduces an innovative solution for classifying flow starvation during square-flow assisted ventilation using convolutional neural network and recurrent neural network models. The majority of existing patient-ventilator asynchrony algorithms [ 37 , 45 ] primarily focus on identifying common forms of asynchronies such as double triggering, ineffective effort, and short- and prolonged cycling. In contrast to previous studies [ 27 – 30 ] employing a binary classification for asynchrony classification, our work adopts a multiclass approach.…”

Section: Discussionmentioning

confidence: 99%

Flow starvation during square-flow assisted ventilation detected by supervised deep learning techniques

de Haro,

Santos-Pulpón,

Telías

et al. 2024

Crit Care

Self Cite

View full text Add to dashboard Cite

Background Flow starvation is a type of patient-ventilator asynchrony that occurs when gas delivery does not fully meet the patients’ ventilatory demand due to an insufficient airflow and/or a high inspiratory effort, and it is usually identified by visual inspection of airway pressure waveform. Clinical diagnosis is cumbersome and prone to underdiagnosis, being an opportunity for artificial intelligence. Our objective is to develop a supervised artificial intelligence algorithm for identifying airway pressure deformation during square-flow assisted ventilation and patient-triggered breaths. Methods Multicenter, observational study. Adult critically ill patients under mechanical ventilation > 24 h on square-flow assisted ventilation were included. As the reference, 5 intensive care experts classified airway pressure deformation severity. Convolutional neural network and recurrent neural network models were trained and evaluated using accuracy, precision, recall and F1 score. In a subgroup of patients with esophageal pressure measurement (ΔPes), we analyzed the association between the intensity of the inspiratory effort and the airway pressure deformation. Results 6428 breaths from 28 patients were analyzed, 42% were classified as having normal-mild, 23% moderate, and 34% severe airway pressure deformation. The accuracy of recurrent neural network algorithm and convolutional neural network were 87.9% [87.6–88.3], and 86.8% [86.6–87.4], respectively. Double triggering appeared in 8.8% of breaths, always in the presence of severe airway pressure deformation. The subgroup analysis demonstrated that 74.4% of breaths classified as severe airway pressure deformation had a ΔPes > 10 cmH2O and 37.2% a ΔPes > 15 cmH2O. Conclusions Recurrent neural network model appears excellent to identify airway pressure deformation due to flow starvation. It could be used as a real-time, 24-h bedside monitoring tool to minimize unrecognized periods of inappropriate patient-ventilator interaction.

show abstract

Section: Discussionmentioning

confidence: 99%

Flow starvation during square-flow assisted ventilation detected by supervised deep learning techniques

de Haro,

Santos-Pulpón,

Telías

et al. 2024

Crit Care

Self Cite

View full text Add to dashboard Cite

show abstract

“…In validating automated characterization of patient–ventilator interaction, consideration should also be given to the shortcomings of the reference. Although esophageal pressure is an established measure for detecting inspiratory patient effort [ 1 , 6 , 10 , 12 ], sometimes amplitudes were very low and hardly recognizable due to low patient activity, muscle weakness, and the volume component in the signal. Temporarily, the pressure signal might also have been disrupted, e.g., by cardiac artifacts or peristalsis.…”

Section: Discussionmentioning

confidence: 99%

“…Beyond that, assisted breaths can be affected by minor asynchrony events, sometimes referred to as dyssynchronies, which occur when triggering or cycling is too early or too late [ 10 ]. Identifying these classes relies on recognizing specific patterns in the airway pressure and flow curves [ 6 , 10 – 12 ] or segmenting the patient’s inspiratory efforts [ 13 , 14 ].…”

Section: Introductionmentioning

confidence: 99%

“…Table 1 summarizes how previous studies have approached this issue. Several authors have attempted to detect and classify asynchronous events in the airway pressure and flow curves both manually [ 2 , 10 , 11 ] and automatically [ 6 , 12 ]. Furthermore, a growing body of literature exists on patient–ventilator asynchrony detection using esophageal pressure [ 1 , 5 , 6 , 10 ], invasively measured diaphragm electromyogram [ 5 , 7 , 11 , 13 ] or noninvasive surface electromyography ( ) [ 4 , 8 , 9 , 14 , 15 ].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Automated characterization of patient–ventilator interaction using surface electromyography

Sauer,

Graßhoff,

Carbon

et al. 2024

Ann. Intensive Care

View full text Add to dashboard Cite

Background Characterizing patient–ventilator interaction in critically ill patients is time-consuming and requires trained staff to evaluate the behavior of the ventilated patient. Methods In this study, we recorded surface electromyography ($$\textrm{sEMG}$$ sEMG ) signals from the diaphragm and intercostal muscles and esophageal pressure ($$P_{\textrm{es}}$$ P es ) in mechanically ventilated patients with ARDS. The sEMG recordings were preprocessed, and two different algorithms (triangle algorithm and adaptive thresholding algorithm) were used to automatically detect inspiratory patient effort. Based on the detected inspirations, major asynchronies (ineffective, auto-, and double triggers and double efforts), delayed and synchronous triggers were computationally classified. Reverse triggers were not considered in this study. Subsequently, asynchrony indices were calculated. For the validation of detected efforts, two experts manually annotated inspiratory patient activity in $$P_{\textrm{es}}$$ P es , blinded toward each other, the $$\textrm{sEMG}$$ sEMG signals, and the algorithmic results. We also classified patient–ventilator interaction and calculated asynchrony indices with manually detected inspirations in $$P_{\textrm{es}}$$ P es as a reference for automated asynchrony classification and asynchrony index calculation. Results Spontaneous breathing activity was recognized in 22 out of the 36 patients included in the study. Evaluation of the accuracy of the algorithms using 3057 inspiratory efforts in $$P_{\textrm{es}}$$ P es demonstrated reliable detection performance for both methods. Across all datasets, we found a high sensitivity (triangle algorithm/adaptive thresholding algorithm: 0.93/0.97) and a high positive predictive value (0.94/0.89) against expert annotations in $$P_{\textrm{es}}$$ P es . The average delay of automatically detected inspiratory onset to the $$P_{\textrm{es}}$$ P es reference was $$-$$ - 79 ms/29 ms for the two algorithms. Our findings also indicate that automatic asynchrony index prediction is reliable. For both algorithms, we found the same deviation of $$0.06\pm 0.13$$ 0.06 ± 0.13 to the $$P_{\textrm{es}}$$ P es -based reference. Conclusions Our study demonstrates the feasibility of automating the quantification of patient–ventilator asynchrony in critically ill patients using noninvasive sEMG. This may facilitate more frequent diagnosis of asynchrony and support improving patient–ventilator interaction.

show abstract

“…19 However, invasive mechanical ventilation was a more common target of ML research, and several studies have attempted to broaden the detection and classification of AE varieties. 20,21,22,23,24 To date, no ML models have been published that directly assess NIV modes of ventilation, either with dedicated NIV machines or ICU ventilators.…”

Section: Discussionmentioning

confidence: 99%

Exploring clinicians' beliefs and practices regarding Non-Invasive Ventilation devices: An international survey study

Benavente,

Robbins,

Fujiuchi

et al. 2023

JMV

View full text Add to dashboard Cite

Introduction Non-invasive ventilation (NIV) has a significant role in supporting patients with respiratory failure with the goal of avoiding mechanical ventilation. Traditionally, NIV has been applied using dedicated NIV-specific devices but over the last decade, newer generation critical care ventilators have updated their capabilities to include NIV options with improved synchrony and leak compensation. No recent trials have compared the efficacy of new generation critical care ventilators to NIV ventilators. The purpose of this study was to evaluate clinicians attitudes and perceptions toward the use of NIV between the dedicated NIV and critical care ventilators. Methods An online survey of clinicians with seven questions regarding their thoughts and experience in using NIV in acute care settings was posted online and promoted through emails and social media. The survey was anonymous and an exemption of consent was obtained from the Institutional Review Board. Analysis of variants (ANOVA) was done for the total responses in each question, followed by multivariate analysis of variants (MANOVA) for responses per occupation. Results 514 responses from 54 countries were recorded. 151 from North America, 109 from South America, 125 from Europe, 97 from Asia, 21 from Africa, and 11 from Australia. 218 responders were physicians, 218 were respiratory therapists, 28 were nurses, and 50 were reported as other professionals (engineers, biomedical technicians). 346 (67.3%) reported using both types of ventilators for NIV, 91 (17.7%) use only NIV -specific devices, and 77 (15%) only use critical care ventilators (P 0.097), responses per occupation (P < 0.001). 290 (56.4%) have automatic synchronization software on either of their ventilators, 113 (22%) do not, while 111 (21.6%) are unsure if they do (P 0.22), with significant variation by occupation (P 0.008). Regarding synchrony, 233 (45.3%) said NIV ventilators are better, and 165 (32.1%) said critical care ventilators are better, while 116 (22.5%) said both are similar (P 0.59) with significant variation by occupation (P 0.04). Regarding leak compensation, 241 (46.9%) said NIV ventilators are better, and 146 (284%) said critical care ventilators are better, while 127 (24.7%) said both are similar (P 0.6) without significant variation by occupation (P 0.07). Regarding the general opinion of superiority, 273 (53.1%) said NIV ventilators are better, 131 (25.5%) said critical care ventilators are better, and 110 (21.4%) said both are similar (P 0.42) without significant variation by occupation (P 0.098). Conclusion Despite the lack of evidence, there is wide variability in opinion with no clear consensus regarding the clinicians’ attitude towards which ventilators are superior to use during NIV, especially according to surveyed occupation.

show abstract

Automated detection and classification of patient–ventilator asynchrony by means of machine learning and simulated data

Cited by 9 publications

References 21 publications

Flow starvation during square-flow assisted ventilation detected by supervised deep learning techniques

Flow starvation during square-flow assisted ventilation detected by supervised deep learning techniques

Automated characterization of patient–ventilator interaction using surface electromyography

Exploring clinicians' beliefs and practices regarding Non-Invasive Ventilation devices: An international survey study

Contact Info

Product

Resources

About