• Home
• Preliminary Program

10:30   Respiration & Pregnancy Chair: Esther Tanck 10:30 15 mins The use of simulated data can improve models for automated detection of patient-ventilator asynchrony Tom Bakkes, Anouk van Diepen, Ashley de Bie, Massimo Mischi, Pierre Woerlee, Simona Turco Abstract: Patient-ventilator asynchrony (PVA) is a hard-to-catch adverse event in mechanically ventilated critically ill patients and is associated with increased mortality [1]. There is a need for automated detection of asynchronies since current solutions are suboptimal. In previous research, a method for the automated detection of patient-ventilator asynchrony (PVA) was developed [2]. Proper validation of these methods is difficult because this requires independent datasets containing accurately labeled PVAs. However, these datasets are difficult to obtain, since labeling the data is labor-intensive, time-consuming, and prone to human errors. In contrast, simulating ventilation data allows for the rapid generation of accurately labeled data and comprehensive datasets. In this study, we investigated the feasibility of obtaining accurately labeled ventilation data from simulations, and the generalizability of the previously developed method to this data. A simulator developed in [3] was used to obtain the simulated data. In total, the simulation generated 64898 breathing cycles. The clinical data was the same data utilized in previous research, which consisted of 15 patients and contained 4275 breathing cycles [2]. The model was trained and tested in four different approaches. First, we performed cross-validation as in [2] with only clinical data. Second, the model was trained on clinical data and tested on simulated data. Third, the model was trained on simulated data and tested on clinical data. Finally, the model was cross-validated on clinical data but the training was supplemented with simulated data. The resulting precision and recall were respectively 97.5% and 98.0% for the first approach, 94.3% and 93.5% for the second approach, 94.5% and 97.8% for the third approach, and 97.6% and 98.3% for the Fourth approach. The overall performance was excellent, with the second and third approaches performing remarkably high, almost similar to the first and fourth approaches in which the algorithms were both trained and tested on clinically labeled data. The second and third approaches confirm that the simulations generate accurate ventilation data for which the detection can be easily generalized to different datasets. This approach with simulated datasets will therefore help to accurately detect PVAs in real-time at the bedside of the patients. [1] C. de Haro et. Al., Intensive care Med Exp. (2019) [2] T. Bakkes et. Al., EMBC. (2020) 10:45 15 mins The effect of prone and supine position ventilation on ventilator induced lung injury Sjeng Quicken, Ulrich Strauch, Frans van de Vosse Abstract: Background: Patients on the intensive care unit (ICU) often require mechanical ventilation (MV). While MV is crucial for ICU patient survival, it often induces ventilator induced lung injury (VILI) which can be life-threatening. VILI mainly develops as a result from inhomogeneous mechanical response of the lung to MV which causes local lung tissue overdistention or collapse. Instead of standard supine position ventilation, patients may be ventilated in prone position to improve MV outcome which has seen a widespread adaptation during the recent COVID pandemic. Research suggests that prone position ventilation can also help minimize VILI. The mechanisms responsible for the positive impact of prone position on VILI development are however largely unknown. In this research a computational respiratory biomechanics model is used to study how prone positioning impacts VILI development. Methods: A realistic airway model was generated using CT data and a respiratory tree growing algorithm. Individual airways were modelled as non-linear resistors. Terminal airways were truncated using lumped-parameter alveolar models. A typical MV pressure curve was prescribed at the trachea, whereas intrapleural pressure distributions for prone and supine position ventilation were estimated from (1, 2). Alveolar overdistention and collapse were assessed by evaluating local alveolar volumetric strains relative to alveolar volume at functional residual capacity. Results: Supine position ventilation resulted in considerably higher strain heterogeneity compared to prone position ventilation. Furthermore, a considerable number of alveoli of the supine position simulation experienced negative strains during the complete respiratory cycle, indicating a tendency for alveolar collapse. In the prone position simulation, alveolar strains were on average higher than those observed in the supine position ventilation and during inhalation, none of the alveolar elements experienced negative strains. Conclusion: Our results suggest that prone position ventilation results in less heterogenous strain conditions and less tendency for alveolar collapse compared to supine position ventilation, which could be a potential mechanism for the benefit of prone position ventilation to combat VILI development. Future research is however still required to further substantiate these finding. 1. A. Kumaresan, et al., Anesthesiology. 128, 1187–1192 (2018). 2. M. H. Tawhai et al., J. Appl. Physiol. 107, 912–920 (2009). 11:00 15 mins An adversarial learning approach to generate mechanical ventilation waveforms for asynchrony detection Liming Hao, Tom Bakkes, Anouk van Diepen, Ashley de Bie Dekker, Pierre Woerlee, Massimo Mischi, Simona Turco Abstract: Patient-ventilator asynchrony (PVA) during mechanical ventilation can lead to pulmonary damage, complications and even increase mortality. A machine learning method was proposed in [1] for detection and diagnosis of PVAs. Proper training of these machine learning detection models requires a large amount of labelled data. However, the availability of clinical data with annotations is limited and simulated data is not ventilator specific. In our research, a new framework based on a generative adversarial network was developed to improve simulated data by learning the ventilator specific artefacts from clinical unlabelled data, with the goal of providing a more powerful training dataset for algorithms as in [1]. The input of the network is the simulated data obtained by the patient-ventilator model developed in [2]. The GAN model consists of a generator and a discriminator. The generator reproduces the characteristics of the clinical data while preserving the annotations of the simulated ones. The role of the discriminator is to distinguish clinical data from generated data, thus pushing the generator to produce more realistic data. Qualitative validation was performed visually by the clinical expert. Our qualitative analysis suggests that GAN model produces more realistic ventilatory waveforms, while maintaining the annotations unchanged. For a quantitative validation of our method, we compared the PVA detection and classification performance of the model in [1] by using as a training dataset the simulated data before and after passing it through GAN model. The TPR and PPV of the PVA detection using generated data were 91.4% and 99.3%, while the ones using simulated data were 94.8% and 98.5%, respectively. In terms of the classification performance, the model trained by generated data performs better than simulated data, with an increase in the F1-score for a majority of the asynchrony types. The ability of the GAN model to produce more realistic simulated data was demonstrated qualitatively and quantitatively. This framework can be integrated with the PVAs detection models to further improve the accuracy of PVA detection and classification, possibly decreasing the risk of pulmonary damage during mechanical ventilation. 11:15 15 mins Development of a new portable headset to monitor energetic load without a mouth mask Charissa Roossien, Bart Verkerke, Michiel Reneman Abstract: A portable headset has been developed to analyze breathing gases and establish the energetic workload of physically active workers more comfortably in comparison to the current golden standard, a mouth mask. This proof-of-concept study aimed to investigate the validity of the portable headset compared to (1) the medical indirect calorimetry method using a mouth mask and (2) oxygen consumption (VO2) estimation based on heart rate, and to explore the user experience of the developed headset system. Fifteen subjects performed a submaximal cycling test twice, once with the headset, and once with the two references systems, a mouth mask and heartrate monitor. Related to indirect calorimetry, good (ICC≥0.72) correlations were observed for the VO2, carbon dioxide production (VCO2) and exhaled volume (Ve). The headset tended to underestimate VO2, VCO2 and Ve at low intensities and to overestimate it at higher intensities. The headset (ICC=0.39) was more valid for estimating VO2 than estimates based on heart rate (ICC=0.11). The subjects preferred the headset over the mouth mask because it was more comfortable, did not hinder communication and had lower breathing resistance. The headset appears to be useable for monitoring the energetic workloads of physically active workers, being more valid than heart rate monitoring and more practical than indirect calorimetry with a mouth mask. The present version is not yet completely valid, but its potential is supported and indicates opportunities for further development and professionalization. This professionalization is on-going in collaboration with industry, aiming to make this headset commercially available. Another design step and further validation studies are needed before implementation in the workplace. There is increasing interest in this breathing-gas analyzing headset to objectively monitor physiological responses of individuals. This system is likely to be of interest as a low-level, comfortable and easy-to-use device for monitoring the physical fitness of subjects in multiple settings, including working, (remote) (occupational) healthcare, rehabilitation and sports settings. It can be carried out in users’ actual environment over longer periods of time. This headset could fill a gap in the existing range of instruments for measuring energy consumption. 11:30 15 mins The effects of advancing gestation on maternal autonomic response Maretha Bester, Rohan Joshi, Massimo Mischi, Judith van Laar, Rik Vullings Abstract: Background: Maternal autonomic adaptation is essential in facilitating the physiological changes that pregnancy necessitates. Insufficient adaptation is linked to complications such as hypertensive diseases of pregnancy. Consequently, tracking autonomic modulation throughout pregnancy could allow for the early detection of emerging deteriorations in maternal health. Autonomic modulation can be longitudinally monitored by assessing heart rate variability (HRV). Yet, changes in maternal HRV corresponding to normally progressing pregnancy remain poorly understood. Current literature focuses on standard HRV features that inform on the activity of the two autonomic branches, often showing conflicting results. Investigating further characteristics of autonomic regulation may offer clarity on autonomic changes during normal pregnancy. One such characteristic is the responsivity of HR to stimuli, which has been shown to be elevated in complicated pregnancies. Subsequently, we investigate whether the increasing stress of healthily advancing gestation alters the maternal autonomic response. Methods: Multiple ECG measurements (≈45 minutes) were obtained longitudinally from 29 healthy pregnant women (range 14-41 weeks of gestation). Maternal autonomic response was assessed with phase rectified signal averaging (PRSA), which graphically shows the rate and magnitude of HR responsivity. Deceleration capacity (DC), which quantifies the response observed in PRSA, was calculated correspondingly. Results were grouped into three gestational age ranges (i.e. under 23 weeks (GA₁), 23 to 32 weeks (GA₂), and over 32 weeks (GA₃)). Friedman’s test, with a Dunn’s post hoc test and Bonferroni correction, and Cohen’s U₁ were performed to determine the significance and effect sizes of differences between groups, respectively. Results: The median and interquartile ranges of the DC were 11.7 (8.3 – 14.9) for GA₁; 9.5 (5.8 – 12.8) for GA₂, and 8.4 (6.2 – 11.7) for GA₃. Changes across groups were significant (p = 0.002), yet the effect sizes were small (U₁ = 0.05, 0.02 and 0.04, respectively). Conclusion: Autonomic responsiveness dampens under the increasing stress of advancing gestation. This downward trend starts before 20 weeks of gestation, i.e. before the timepoint after which complications are typically diagnosed. Subsequently, longitudinally tracking maternal autonomic response with PRSA may aid in the early detection of complications. 11:45 15 mins Automated prediction of the time to labor from the electrohysterogram Ivar de Vries, Judith van Laar, Beatrijs van der Hout-van der Jagt, Rik Vullings Abstract: Preterm labor (i.e. labor before 37 weeks of gestation) is a major cause of neonatal morbidity and mortality. Early detection of preterm birth allows for adequate treatment and could therefore improve neonatal outcomes. Such treatment might consist of the admission of medication to suppress labor and accelerate fetal lung maturation. Since these treatments are dependent on gestational age and have sensitive timing, clinical practice might benefit from a method for accurate detection of time to labor. Using time windows of the uterine electrical activity (electrohysterogram) during pregnancy, this work aims at developing a neural-network based method for the prediction of time until (spontaneous) labor. For this purpose, a mean-variance estimator is trained on features extracted from the electrohysterogram. Kernel-density estimation is used to recombine the predictions and generate a single probability distribution. Given the high incidence of induced labor (>50%), effort is made to include these cases with contaminated labels during training. The presented method will be evaluated as a screening tool for birth at different relevant gestational ages. We aim to apply this tool towards the end of the second trimester of pregnancy. Evaluated on a combination of the open-access Term-Preterm EHG and Icelandic 16-lead Electrohysterogram databases, a mean absolute error of 12.1 days, with an estimated standard deviation (i.e. uncertainty) of 12.3 days was found. This uncertainty in the model predictions is lowered to 10.4 days by using a custom loss, which allows us to include samples with induced labor during training.

end %-->