Introduction: Lung protective ventilation requires low tidal volume and suitable positive end-expiratory pressure (PEEP). To date, few methods have been accepted for clinical use to set the appropriate PEEP. The aim of this study was to test the feasibility of PEEP titration guided by ventilation homogeneity using the global inhomogeneity (GI) index based on electrical impedance tomography (EIT) images. Methods: In a retrospective study, 10 anesthetized patients with healthy lungs mechanically ventilated under volume-controlled mode were investigated. Ventilation distribution was monitored by EIT. A standardized incremental PEEP trial (PEEP from 0 to 28 mbar, 2 mbar per step) was conducted. During the PEEP trial, "optimal" PEEP level for each patient was determined when the air was most homogeneously distributed in the lung, indicated by the lowest GI index value. Two published methods for setting PEEP were included for comparison based on the maximum global dynamic compliance and the intra-tidal compliance-volume curve. Results: No significant differences in the results were observed between the GI index method (12.2 ± 4.6 mbar) and the dynamic compliance method (11.4 ± 2.3 mbar, P > 0.6), or between the GI index and the compliance-volume curve method (12.2 ± 4.9 mbar, P > 0.6). Conclusions: According to the results, it is feasible and reasonable to use the GI index to select the PEEP level with respect to ventilation homogeneity. The GI index may provide new insights into the relationship between lung mechanics and tidal volume distribution and may be used to guide ventilator settings.

The study objective was to compare titration of positive end-expiratory pressure (PEEP) with electrical impedance tomography (EIT) and with ventilator-embedded pressure–volume loop in severe acute respiratory distress syndrome (ARDS). We have designed a prospective study with historical control group. Twenty-four severe ARDS patients (arterial oxygen partial pressure to fractional inspired oxygen ratio, PaO2/FiO2 < 100 mmHg) were included in the EIT group and examined prospectively. Data from another 31 severe ARDS patients were evaluated retrospectively (control group). All patients were receiving medical care under identical general support guidelines and protective mechanical ventilation. The PEEP level selected in the EIT group was the intercept point of cumulated collapse and overdistension percentages curves. In the control group, optimal PEEP was selected 2 cmH2O above the lower inflection point on the static pressure–volume curve. Patients in the EIT group were younger (P < 0.05), and their mean plateau pressure was 1.5 cmH2O higher (P < 0.01). No differences in other baseline parameters such as APACHE II score, PaO2/FiO2, initial PEEP, driving pressure, tidal volume, and respiratory system compliance were found. Two hours after the first PEEP titration, significantly higher PEEP, compliance, and lower driving pressure were found in the EIT group (P < 0.01). Hospital survival rates were 66.7% (16 of 24 patients) in the EIT group and 48.4% (15 of 31) in the control group. Identical rates were found regarding the weaning success rate: 66.7% in the EIT group and 48.4% in the control group. In severe ARDS patients, it was feasible and safe to guide PEEP titration with EIT at the bedside. As compared with pressure–volume curve, the EIT-guided PEEP titration may be associated with improved oxygenation, compliance, driving pressure, and weaning success rate. The findings encourage further randomized control study with a larger sample size and potentially less bias in the baseline data. Trial Registration NCT03112512

The electrical impedance tomography (EIT)-based global inhomogeneity (GI) index was introduced to quantify tidal volume distribution within the lung. Up to now, the GI index was evaluated for plausibility but the analysis of how it is influenced by various physiological factors is still missing. The aim of our study was to evaluate the influence of proportion of open lung regions measured by EIT on the GI index.

Convolutional neural networks (CNNs) have become a useful tool for a wide range of applications such as text classification. However, CNNs are not always sufficiently accurate to be useful in certain applications. The selection of activation functions within CNN architecture can affect the efficacy of the CNN. However, there is limited research regarding which activation functions are best for CNN text classification. This study tested sixteen activation functions across three text classification datasets and six CNN structures, to determine the effects of activation function on accuracy, iterations to convergence, and Positive Confidence Difference (PCD). PCD is a novel metric introduced to compare how activation functions affected a network’s classification confidence. Tables were presented to compare the performance of the activation functions across the different CNN architectures and datasets. Top performing activation functions across the different tests included the symmetrical multi-state activation function, sigmoid, penalised hyperbolic tangent, and generalised swish. An activation function’s PCD was the most consistent evaluation metric during activation function assessment, implying a close relationship between activation functions and network confidence that has yet to be explored.

Introduction: Limiting the energy transfer between ventilator and lung is crucial for ventilatory strategy in acute respiratory distress syndrome (ARDS). Part of the energy is transmitted to the viscoelastic tissue components where it is stored or dissipates. In mechanically ventilated patients, viscoelasticity can be investigated by analyzing pulmonary stress relaxation. While stress relaxation processes of the lung have been intensively investigated, non-linear interrelations have not been systematically analyzed, and such analyses have been limited to small volume or pressure ranges. In this study, stress relaxation of mechanically ventilated lungs was investigated, focusing on non-linear dependence on pressure. The range of inspiratory capacity was analyzed up to a plateau pressure of 45 cmH₂O. Methods: Twenty ARDS patients and eleven patients with normal lungs under mechanical ventilation were included. Rapid flow interruptions were repetitively applied using an automated super-syringe maneuver. Viscoelastic resistance, compliance and time constant were determined by multiple regression analysis using a lumped parameter model. This same viscoelastic model was used to investigate the frequency dependence of the respiratory system's impedance. Results: The viscoelastic time constant was independent of pressure, and it did not differ between normal and ARDS lungs. In contrast, viscoelastic resistance increased non-linearly with pressure (normal: 8.4 (7.4-11.9) [median (lower - upper quartile)] to 35.2 (25.6-39.5) cmH₂O·sec/L; ARDS: 11.9 (9.2-22.1) to 73.5 (56.8-98.7)cmH₂O·sec/L), and viscoelastic compliance decreased non-linearly with pressure (normal: 130.1(116.9-151.3) to 37.4(34.7-46.3) mL/cmH₂O; ARDS: 125.8(80.0-211.0) to 17.1(13.8-24.7)mL/cmH₂O). The pulmonary impedance increased with pressure and decreased with respiratory frequency. Conclusions: Viscoelastic compliance and resistance are highly non-linear with respect to pressure and differ considerably between ARDS and normal lungs. None of these characteristics can be observed for the viscoelastic time constant. From our analysis of viscoelastic properties we cautiously conclude that the energy transfer from the respirator to the lung can be reduced by application of low inspiratory plateau pressures and high respiratory frequencies. This we consider to be potentially lung protective.