An open-source software for automatic calculation of ...

Respiratory Physiology & Neurobiology 192 (2014) 1–6

Contents lists available at ScienceDirect

Respiratory Physiology & Neurobiology journal homepage: www.elsevier.com/locate/resphysiol

An open-source software for automatic calculation of respiratory parameters based on esophageal pressure Louis Mayaud a,b,c,d,∗ , Michèle Lejaille a,b,c , Hélène Prigent a,b,c , Bruno Louis e , Brigitte Fauroux f , Frédéric Lofaso a,b,c a

INSERM, Centre d’Investigation Clinique et d’Innovation technologique (CICIT), UMR805, Garches, France EA4497, Université Versailles Saint-Quentin (UVSQ), Versailles, France c Hôpital Raymond Poincaré, APHP, Garches, France d Institute of Biomedical Engineering (IBME), University of Oxford, Oxford, UK e INSERM U955, 94000 Créteil, France f Pediatric Pulmonology, Hopital Armand Trousseau, AP-HP, and Pierre et Marie Curie University, Paris, France b

a r t i c l e

i n f o

Article history: Accepted 12 November 2013 Keywords: Acute respiratory failure Respiratory function Esophageal pressure Pressure time product PEEPi

a b s t r a c t Purpose: We have developed a software that automatically calculates respiratory effort indices, including intrinsic end expiratory pressure (PEEPi) and esophageal pressure–time product (PTPeso). Materials and methods: The software first identifies respiratory periods. Clean signals are averaged to provide a reference mean cycle from which respiratory parameters are extracted. The onset of the inspiratory effort is detected automatically by looking backward from the onset of inspiratory flow to the first point where the esophageal pressure derivative is equal to zero (inflection point). PEEPi is derived from this point. Twenty-three recordings from 16 patients were analyzed with the algorithm and compared with experts’ manual analysis of signals: 15 recordings were performed during spontaneous breathing, 1 during non-invasive mechanical ventilation, and 7 under both conditions. Results: For all values, the coefficients of determinations (r2 ) exceeded 0.94 (p < 0.001). The bias (mean difference) between PEEPi calculated by hand and automatically was −0.26 ± 0.52 cm H2 O during spontaneous breathing and the precisions (standard deviations of the differences) was 0.52 cm H2 O with limits of agreement of 0.78 and −1.30 cm H2 O. The mean difference between PTPeso calculated by hand and automatically was −0.38 ± 1.42 cm H2 O s/cycle with limits of agreement of 2.46 and −3.22 cm H2 O s/cycle. Conclusions: Our program provides a reliable method for the automatic calculation of PEEPi and respiratory effort indices, which may facilitate the use of these variables in clinical practice. The software is open source and can be improved with the development and validation of new respiratory parameters. © 2013 Elsevier B.V. All rights reserved.

1. Introduction Recently, Brochard et al. (2012b) have suggested a list of respiratory parameters to be monitored in critically ill patients. In

Abbreviations: Ccw, chest wall compliance; CLdyn, dynamic lung compliance; F, female; FRC, functional residual capacity; EEV, end-expiratory volume; PEEPi, intrinsic positive end expiratory pressure; ID, identification; IVC, inspiratory vital capacity; M, male; MV, mechanical ventilation; NIVM, non-invasive mechanical ventilation; Pmask, mask pressure; Pdi, transdiaphragmatic pressure; Peso, oesophageal pressure; Pga, gastric pressure; Ptp, transpulmonary pressure; PTPdi, transdiaphragmatic product time product; PV, pressure–volume; RSR, respiratory system resistance; SB, spontaneous breathing; Ti, inspiratory time; TLC, total lung capacity; VT, tidal volume; RV, residual volume; WOB, work of breathing. ∗ Corresponding author at: CICIT, Hôpital Raymond Poincaré, 104, bd Boulevard Raymond Poincaré, 92380 Garches, France. Tel.: +33 650663491; fax: +33 147104633. E-mail address: [email protected] (L. Mayaud). 1569-9048/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.resp.2013.11.007

particular, assessments of respiratory effort in patients with respiratory failure are useful, and even more in patients requiring mechanical ventilation. The esophageal pressure time product (PTPeso) is considered as one of the most useful variables for quantifying respiratory muscle effort (Baydur et al., 1982b; Sassoon et al., 1991; Tobin and Laghi, 1998) and this variable has been broadly used in recent literature (Carr et al., 2012; Jubran, 1999; Man et al., 2012). PTPeso is calculated as the time integral of the difference between the oesophageal pressure (Peso) and the estimated recoil pressure of the chest wall during inspiration (Baydur et al., 1982a; Jubran, 1999). However, this measurement requires the determination of the onset of the inspiratory effort and the estimation of the intrinsic positive end expiratory pressure (PEEPi) to determine the precise onset of inspiration to the chest wall relaxation line (Baydur et al., 1982a; Chu and Han, 2008; Sassoon et al., 1991). Few commercially available systems for monitoring these indices are available and it

2

L. Mayaud et al. / Respiratory Physiology & Neurobiology 192 (2014) 1–6

is unknown if these systems are able to adequately estimate PEEPi (Baydur et al., 1982a; Sassoon et al., 1991; Tobin and Laghi, 1998). We have developed a MATLAB-based standalone program, which is designed to automatically detect the onset of the inspiratory effort and calculate the inspiratory effort indices taking into account PEEPi (Tobin and Laghi, 1998). The aim of our study was to validate this program by comparing automated versus manual (expert) calculation of PEEPi and PTPeso on different patients’ recordings in order to allow an online access of the validated software. The software and its documentation are published online (http://code.google.com/p/respmat/), where source code is available and can easily be improved with the development and validation of new respiratory parameters. 2. Methods 2.1. Patients We used data previously recorded from different studies at Raymond Poincaré hospital on adult patients with neuromuscular diseases and at Trousseau hospital on pediatric patients with neuromuscular diseases, cystic fibrosis and obstructive sleep apnea. Studies were all approved by Institutional Review Board (IRB, Saint Antoine Faculty and Île de France Saint Germain en Laye: NCT01113255) and written informed consent had been obtained from all patients and his/her parents when applicable. Among the 86 patients who were initially included in the previous clinical studies and for whom the transdiaphragmatic recording was available, only 14 had a manually detectable PEEPi above 1 cm H2 O and were thereby included in the study. In addition, one adult and one child without PEEPi had been randomly selected, in order to check the absence of false positive with the automated method of PEEPi measurement. Finally, 23 recordings of 16 patients (11 neuromuscular adult patients and 5 pediatric patients) were analyzed: 15 of them were recorded during spontaneous breathing, 8 of them were recorded during non-invasive mechanical ventilation (NIMV), and 7 patients were recorded during both conditions. 2.2. Measurements All the measurements were performed with the patient in 45◦ semi-recumbent position in order to avoid leak during NIMV (Antonello et al., 2013). First, all patients (except patient number 13) were recorded during spontaneous breathing (5 min) and then under NIMV after a 15-min-long stabilization period. Patients’ ventilator parameters were used when available, or were determined according to standard clinical practice (Fauroux et al., 2004): pressure support ventilation (PSV) was initialized at 8 cm H2 O and gradually increased by 1 cm H2 O until maximum patient comfort was met, which was defined by the disappearance (or maximal decrease) of shortness of breath, as perceived by the patients (Make et al., 1998). The pressurization rate and trigger level were determined following a similar procedure and the backup respiratory (Contal et al., 2013) rate was set up at 10 cycles per minute (Fauroux et al., 2004). Mask pressure (Pmask) was measured with a differential pressure transducer (MP 45 model, Validyne ± 100 cm H2 O, Northridge, CA) on the mask and was only used for the analysis of mechanically ventilated patients. The pneumotachograph was positioned between the mask and the ventilator circuit Y-piece. Flow was measured using a Fleisch #1 pneumotachograph (Lausanne, Switzerland) connected to a pressure transducer (MP 45 model, Validyne ± 2 cm H2 O). Peso and gastric (Pga) pressures were measured using a 2.1 mm external diameter probe with two integrated pressure transducers (Gaeltec, Dunvegan, Isle of Skye, UK) inserted per-nasally after careful local anesthesia (lidocaine 2%,

Astra Zeneca, Rueil-Malmaison, France). After the calibration of the two pressure transducers using a 10 cm H2 O water-column, the catheter was advanced gently until the distal tip was in the stomach and the proximal pressure transducer in the middle portion of the esophagus. Appropriate placement of the Peso transducer was assessed with the occlusion technique (Baydur et al., 1982a). Adequate placement of the Pga transducer was ascertained by gentle manual pressure on the patient’s abdomen to observe fluctuations in Pga, which should be absent or less present on the Peso trace. In addition we checked that, when the patient was given water to drink, the tracing showed a sharp rise in Peso resulting from muscular contraction of the esophagus, without any concomitant modification of Pga. All the signals were digitized at 128 Hz and sampled for analysis using an analogical/digital acquisition system (MP 100, Biopac Systems, Goletta, CA), run on a PC computer with Acknowledge software (version 3.9). 2.3. Automated calculation Respiratory signals were filtered with a 5th-order band-pass Chebyshev filter (between 0.05 Hz and 5 Hz). Candidates for onset of inspiratory flow where identified with zero crossing algorithm applied to the flow signal. A zero crossing algorithm identifies two consecutive samples of different signs. Respiratory cycles were successively excluded from further analysis according to the following criteria: 1. Period was less than 1 s (artifacts in zero crossing identification); 2. Period length departing from more than 3 standard deviation from mean period time (respiratory cycles that are not representative of the average pattern); 3. Period with at least one data point in Peso, Pga or Pmask that departed from more than 3 standard deviation from the mean value of the signal (respiratory cycles with noisy signals); 4. Period with expiratory volumes having more than 30% relative difference compared to the inspiratory volume (respiratory cycle with a flow leak). The average period was computed from remaining cycles for all signals Pmask, Peso, Pga and flow and new signals were derived: • Volume was the integration of flow with time; • Transdiaphragmatic pressure (Pdi) was defined as Pga minus Peso; • Ptp was defined as Pmask minus Peso. The compliance of the chest wall (Chu and Han, 2008) (Ccw) was estimated as 4% of the theoretical inspiratory vital capacity (IVC) (Quanjer et al., 1995) divided by 1 cm H2 O. The IVC was computed from height (h in cm), age (a in years) and gender as follow: • Adult (18 > age > 77 years) ◦ Men: IVC = 6.10h − 0.028a − 4.65 ◦ Women: IVC = 4.66h − 0.026a − 3.28 • Teenagers (age < 18 years and height > 150 cm) ◦ Boys: IVC = 8.4h − 9.9 ◦ Girls: IVC = 5.0h − 4.5 • Children (age < 18 years and height < 150 cm) ◦ Boys: IVC = 5.70h − 5.26 ◦ Girls: IVC = 5.50h − 5.39 The volume signal was linearly corrected for air leaks, to provide similar inspiratory and expiratory volumes averaged on the entire record. The PEEPi was defined as the positive recoil pressure of the respiratory system at end-expiration and is therefore equal to Peso at the beginning of the inspiratory effort minus Peso at the

L. Mayaud et al. / Respiratory Physiology & Neurobiology 192 (2014) 1–6

3

dV/dt (L/s)

Ccw

CLdyn

TLC

Pdi (cmH2O)

Volume (L)

PEEPi

EEV

PTPdi

Peso (cmH2O)

Voffset

Pcw

FRC

Peso - Pmask (cmH2O)

PTPeso PEEPi E

T (s)

RV

Ti

0

Fig. 1. (Left) Campbell diagram with pressure–volume curve (dark blue), chest wall compliance line (red) and dynamic lung compliance (green). (Right) time plots for airflow (top), transdiaphragmatic (middle) and esophageal (bottom) pressures. The beginning of the inspiratory effort and the actual inspiration are indicated on both side of the figure with the letter E and A, respectively. Similarly the end of the inspiration and beginning of the expiration is noted B. Legend: Ccw, chest wall compliance; CLdyn, dynamic lung compliance; EEV, end-expiratory volume; FRC, functional residual capacity; Pdi, transdiaphragmatic pressure; PEEPi, intrinsic positive end expiratory pressure; Peso, oesophageal pressure; PTPdi, transdiaphragmatic product time product; RV, residual volume; Ti, inspiratory time; TLC, total lung capacity.

beginning of the inspiratory flow. We then used these signals to build a Campbell pressure–volume diagram (Brochard et al., 2012a) from which works of breathing were derived as seen in Fig. 1 and detailed below. In particular, we fitted dynamic lung compliance (CLdyn) and chest wall compliance (Ccw) lines to the beginning and the end of the respiratory cycle, respectively. The intersection of the two compliance lines, defines the functional residual capacity (FRC), i.e. the relaxation volume and defines a new volume (Voffset ), which accounts for dynamic over-inflation above relaxation volume (Tuxen and Lane, 1987). In the example of hyperinflation shown in Fig. 1 (left-hand side), the difference between the end expiratory volume (EEV) and FRC noted Voffset is derived from simple Euclidean equations under the geometrical constraints presented in Fig. 2 giving: a + b = PEEPi VOffset b

(2)

Volume (L)

|CLdyn | =

(1)

Ccw

CLdyn PEEPi

c

d

Voffset

Peso - Pmask (cmH2O)

a

b

Fig. 2. Geometry identified to determine the offset volume from Campbell diagram. Chest wall compliance line (red) and dynamic lung compliance (green) are fitted to beginning of inspiratory effort and inspiration, respectively. Their intersection point (I) defines the offset volume, which equation can be derived from simple Euclidian geometry equations.

Ccw =

VOffset a

(3)

Summing (2) + (3) and substituting (1) gives: Ccw + |CLdyn | =

PEEPi · VOffset a·b

(4)

where a · b can be identified from Eqs. (2) and (3), leading to: a·b=−

2 VOffset

Ccw .|CLdyn |

(5)

Finally, substitution of (5) in (4) leads to the final equation of the offset volume: VOffset =

PEEPi · Ccw · |CLdyn | |CLdyn | + Ccw

In theory, the dynamic lung compliance CLdyn is defined as the ratio of change in volume to change in Ptp between instant of zeroflow within the same breath and should therefore be considered as negative. To avoid confusion, we present equations using its absolute value. The point where inspiratory effort starts simply and automatically as the first point, backward from the beginning the inspiratory flow, where the first derivative of the Peso is equal to zero (when the curve shows its first inflection point). Absence of active expiratory effort was confirmed by the absence of Pga increase during expiration whereas abdominal section continued to decrease (Fauroux et al., 2004). The PTPeso is defined by as the area (time integral) between the Peso and the estimated recoil pressure of the chest wall from the beginning of the inspiratory effort to the end of the inspiratory flow (Tobin and Laghi, 1998) as indicated by the red area on the right side of Fig. 1. In addition to this, the decrease in Peso before flow onset attributed to inspiratory activity needed to overcome the threshold load imposed by dynamic hyperinflation and was therefore taken in account. Likewise, the transdiaphragmatic PTP (PTPdi) is defined as the area under the Pdi signal from the onset of its positive deflection to the end of the inspiratory flow, which is represented by the green area in Fig. 1.

4 Table 1 Description of patients included in the study and available recording modes used in this study (mode). Patients aged less than 18 are pediatrics patients (# 1 to 5). All parameters presented in the right-hand part of the table were extracted with the automated software using Spontaneous Breathing (SB) recording (except for patient # 13). All patients for whom a mechanical ventilation (MV) recording was available were non-invasively ventilated by using the pressure support mode. Patient’s number (#)

2 3

Pressure support inspiratory–expiratory pressure (cm H2 O)

Gender

Age (years)

Weight (kg)

Height (cm)

RSR (cm H2 O/L/s)

PEEPi (cmH2 O)

CLdyn (L/cmH2 O)

Swing Peso (cmH2 O)

Pathology

SB MV SB MV SB

16-0

M

14

35

145

15.8

2.8

0.037

−18.3

Cystic fibrosis

16-0

M

17

52

170

14.9

1.4

0.042

−20.3

Myopathy

8-0

M

15

44

160

8.3

0.0

0.037

−8.1

16-0

F

15

38

158

29.1

3.0

0.028

−16.7

Spinal muscular atrophy Cystic fibrosis

11-5

M

12

30

133

20.7

1.1

0.073

−10.7

Laryngeal diastema

6 7

SB MV SB MV SB SB

N.A. N.A.

F M

74 44

93 74

170 166

23.1 9.1

4.1 4.3

0.056 0.058

−18.2 −15.8

8 9

SB SB

N.A. N.A.

M F

57 68

79 63

178 160

38.5 12.5

6.5 2.9

0.037 0.043

−27.0 −13.3

SB SB SB MV MV

N.A. N.A. 13-0

M M M

81 29 66

79 31 75

174 168 185

8.4 8.8 7.5

2.6 0.0 3.5

0.074 0.042 0.618

−9.1 −9.3 −9.0

14-0

M

28

67

172

2.0

0.9

0.067

−8.8

SB MV SB MV SB

13-0

M

42

140

177

11.4

5.0

0.074

−15.6

12-0

M

38

98

183

6.7

1.6

0.222

−6.6

N.A.

M

73

75

175

5.2

0.1

0.083

−12.5

Pompe’s disease Unilateral phrenic neuropathy Thoracic trauma Mixed respiratory failure with ICU-acquired paresis Tracheomalacia Duchenne’s myopathy Steinert myotonic dystrophy Steinert myotonic dystrophy Steinert myotonic dystrophy Steinert myotonic dystrophy Amyotrophic lateral sclerosis

4 5

10 11 12 13 14 15 16

Abbreviations: F, female; ICU, intensive care unit; PEEPi, intrinsic post end expiratory pressure; LC, lung compliance; M, male; MV, mechanical ventilation; N.A., not applicable; Peso, oesophageal pressure; RSR, respiratory system resistance; SB, spontaneous breathing. RSR was calculated as mean resistive pressure (derived from the inspiratory WOB in Fig. 1) divided by mean inspiratory flow.

L. Mayaud et al. / Respiratory Physiology & Neurobiology 192 (2014) 1–6

1

Mode

L. Mayaud et al. / Respiratory Physiology & Neurobiology 192 (2014) 1–6

The different works of breathing (WOB) are derived from the Campbell diagram as seen in Fig. 1: • Wres , the inspiratory resistive work is defined by the area delimited by the PV curve (blue) on the left and the CLdyn line (green) on the right (1); • Wel , the elastic work is defined by the green area in the triangle delimited by A, B and C; • WPEEPi , the additional elastic work imposed by the PEEPi defined by the blue trapeze delimited by A, C, D and E; • Wexp , the expiratory work is defined by the area delimited by the PV curve (blue) on the right and the Ccw line (red) on the left (2).

For mechanically ventilated patients (Pmask greater than 3 cm H2 O), it was not possible to distinguish between the patient’s effort and the mechanical ventilation input. Therefore, the area between the pressure–volume curve during inspiration (A to B in Fig. 1) and the Ccw line (ED in red in Fig. 1) is computed and noted W, the total work.

2.4. Manual calculation To limit the burden of manual comparison and in accordance with clinical practice, experts (FL and ML) identified, for each patient, a single respiratory cycle. The identification of a single cycle allowed the validation of the automated technique based on the exact same waveform, while the clinically relevant information would obviously be spread over several clean cycles accounting for natural breathing variability. Based on this cycle, experts manually measured inspiratory time (Ti), total cycle time (Ttot), tidal volume (VT) by integrating the inspiratory flow, PEEPi and PTPeso as previously described (Sassoon et al., 1991). Signals were not pre-filtered to validate the use of filtering in the automated analysis.

2.5. Comparison Automated versus hand-calculations were compared using Bland–Altman plot when considered different (>1% in only one condition).

3. Results The population studied is described in Table 1. For each recording, differences between the two measurement methods were found to be less than 1% for Ti, Ttot and VT. As expected, there was a significant positive correlation between PEEPi calculated by hand and automatically (r2 = +0.931, p < 0.001). The left panel of Fig. 3 shows the Bland and Altman plot of the difference between PEEPi calculated by hand and automatically with a mean difference between of −0.26 ± 0.52 cm H2 O with limits of agreement of 0.78 and −1.30 cm H2 O. There was a positive correlation between PTPeso calculated by hand and automatically (r2 = +0.977, p < 0.0001). The right panel of Fig. 3 shows the Bland and Altman plot of the difference between PTPeso measured by hand and automatically with a mean difference between of −0.38 ± 1.42 cm H2 O s/cycle with limits of agreement of 2.46 and −3.22 cm H2 O s/cycle. 4. Discussion This study validates an automated method using filters to determine the onset of inspiratory effort and subsequently the PEEPi measurement and PTPeso calculation. The validation consisted in comparing manual versus automated calculation of these parameters during spontaneous breathing and NIMV in pediatric and adult patients with various diseases. Although our program was able to calculate the WOB, we did not compare this parameter against its manual calculation because validation of PTPeso will provide good confidence in the former. Indeed, Ti error only affects PTPeso calculation, whereas all the error measurements that affect WOB, like VT and PEEPi, will also automatically affect PTPeso. Moreover, WOB can be measured either by planimetry of pressure–volume curve or by integration of the pressure-flow product, and there is no reason to observe differences between the two methods since VT and PEEPi are comparable. We proposed a Pga recording in our program for two reasons. First, this allows the calculation of the PTPdi (Sassoon et al., 1991). This is useful to distinguish the implication of the diaphragm from other inspiratory muscles in a single patient, which is particularly helpful for the critically ills. Second, it allows the detection of PEEPi due to expiratory muscle activity (Lessard et al., 1995). In this case, external PEEP will add to the expiratory work whereas when PEEPi is due to hyperinflation, external PEEP reduces the inspiratory effort.

5 PEEPi(H) – PEEPi(A) (cmH2O)

3 2 1

Mean+2SD = +0.78

0

Mean = -0.26

-1

Mean-2SD = -1.30

-2 -3 -4 0

1

2

3

4

5

Average PEEPi (cmH2O)

6

7

8

PTPeso(H) – PTPeso(A) (cmH2O.sec/cycle)

10

4

-5

5

5 Mean+2SD = +2.46 Mean = -0.38

0

Mean-2SD = -3.22

-5

-10

0

5

10

15

20

25

30

35

Average PTPeso (cmH2O.sec/cycle)

Fig. 3. Left panel: the Bland and Altman plot of the difference between PEEPi calculated by hand (H) and PEEPi automatically calculated (A). Right Panel: the Bland and Altman plot of the difference between the oesophageal pressure time product (PTPeso) calculated by hand (H) and automatically (A). (䊉) Patients during spontaneous breathing, () patients under non-invasive ventilation.

6

L. Mayaud et al. / Respiratory Physiology & Neurobiology 192 (2014) 1–6

The software presented here is able to automatically detect PEEPi (and other validated parameters) in various conditions (spontaneous breathing and NIMV) and patients (adult and pediatric diseases). However, because algorithms can hardly ever encompass the broad variety of signals and noises, we give the user the possibility to adjust several parameters: levels of artifact rejection and onset of respiratory effort location. Most importantly, we provide a graphical user interface that will facilitate the monitoring and clinical use of respiratory effort indexes (see screen Additional File 1 for screenshot with waveforms). While the importance of respiratory parameters to monitor critically ill patients has been recently stressed (Brochard et al., 2012a), they have mainly remained confined to the research environment. Commercial systems that are compact, portable, robust, and relatively easy to use, are now available (Fauroux et al., 2004). However, there has been so far no study comparing the automated versus manual calculations of indices of respiratory effort in patients with PEEPi (Tobin and Laghi, 1998). Supplementary material related to this article can be found, in the online version, at http://dx.doi.org/10.1016/j.resp.2013.11.007. Last but not least, the source code of this program is made available to readers through an open source platform. We believe that this will allow the quick development and validation of new parameters extracted from flow and pressure recordings during spontaneous breathing or mechanical ventilation. For instance, the software can be easily improved to explore hypothesis introduced in most recent studies in critical care or other field or respiratory medicine (Carr et al., 2012; Chervin et al., 2012; Contal et al., 2013; Heulitt et al., 2012; Man et al., 2012). Limitation of the automated analysis includes the lack of signal quality index, requiring the presence of an expert during data analysis. Similarly, while an attempt has been made to account for possible leaks during the signal recording, special care should be given to data acquisition. Limitation of the algorithm validation lies in the selection of a single representative cycle for comparison of the two techniques, which was imposed by the tedious and time consuming manual analysis of the data by the experts. Finally, this study is restricted to patients with neuromuscular diseases and pediatric patients with cystic fibrosis and obstructive sleep apnea, without patients under invasive mechanical ventilation and other type of diseases such as chronic obstructive pulmonary disease. Therefore further studies are needed to confirm the effectiveness of this program for these other populations. 5. Conclusion We have developed an algorithm for analysis of flow and Peso recordings that automatically computes PTPeso, PEEPi in addition to several other respiratory parameters. This algorithm has been compared to manual analysis of these parameters and shows acceptable agreement with expert measurement. The software is made publicly available in an open-source platform with a graphical user interface, which hopefully will ease the monitoring

of lung function in critically ill patients. In addition to this, simple modification of the source code will allow clinicians to explore and validate new respiratory parameters. References Antonello, N., Paolo, B., Cornelius, B., Gianluca, F., Agata, L., Bruna, G., 2013. Noninvasive ventilation in the treatment of sleep-related breathing disorders: concise clinical review. Journal of General Practice 1, 2. Baydur, A., Behrakis, P., Zin, W., Jaeger, M., Milic-Emili, J., 1982a. A simple method for assessing the validity of the esophageal balloon technique. American Review of Respiratory Disease 126, 788–791. Baydur, A., Behrakis, P.K., Zin, W.A., Jaeger, M., Milic-Emili, J., 1982b. A simple method for assessing the validity of the esophageal balloon technique. American Review of Respiratory Disease 126, 788–791. Brochard, L., Martin, G.S., Blanch, L., Pelosi, P., Belda, F.J., Jubran, A., Gattinoni, L., Mancebo, J., Ranieri, V.M., Richard, J.-C.M., 2012a. Clinical review: respiratory monitoring in the ICU – a consensus of 16. Critical Care 16, 219. Brochard, L., Martin, G.S., Blanch, L., Pelosi, P., Belda, F.J., Jubran, A., Gattinoni, L., Mancebo, J., Ranieri, V.M., Richard, J.C., Gommers, D., Vieillard-Baron, A., Pesenti, A., Jaber, S., Stenqvist, O., Vincent, J.L., 2012b. Clinical review: respiratory monitoring in the ICU – a consensus of 16. Critical Care 16, 219. Carr, G.E., Mokhlesi, B., Gehlbach, B.K., 2012. Acute cardiopulmonary failure from sleep-disordered breathing. Chest 141, 798–808. Chervin, R.D., Ruzicka, D.L., Hoban, T.F., Fetterolf, J.L., Garetz, S.L., Guire, K.E., Dillon, J.E., Felt, B.T., Hodges, E.K., Giordani, B.J., 2012. Esophageal pressures, polysomnography, and neurobehavioral outcomes of adenotonsillectomy in children. CHEST Journal 142, 101–110. Chu, M.W., Han, J.K., 2008. Introduction to pulmonary function. Otolaryngologic Clinics of North America 41, 387–396. Contal, O., Adler, D., Borel, J.C., Espa, F., Perrig, S., Rodenstein, D., Pepin, J.L., Janssens, J.P., 2013. Impact of different backup respiratory rates on the efficacy of noninvasive positive pressure ventilation in obesity hypoventilation syndrome: a randomized trial. Chest 143, 37–46. Fauroux, B., Nicot, F., Essouri, S., Hart, N., Clement, A., Polkey, M.I., Lofaso, F., 2004. Setting of noninvasive pressure support in young patients with cystic fibrosis. European Respiratory Journal: Official Journal of the European Society for Clinical Respiratory Physiology 24, 624–630. Heulitt, M.J., Clement, K.C., Holt, S.J., Thurman, T.L., Jo, C.-H., 2012. Neurally triggered breaths have reduced response time, work of breathing, and asynchrony compared with pneumatically triggered breaths in a recovering animal model of lung injury. Pediatric Critical Care Medicine 13, e195–e203. Jubran, A., 1999. Advances in respiratory monitoring during mechanical ventilation. Chest 116, 1416–1425. Lessard, M.R., Lofaso, F., Brochard, L., 1995. Expiratory muscle activity increases intrinsic positive end-expiratory pressure independently of dynamic hyperinflation in mechanically ventilated patients. American Journal of Respiratory and Critical Care Medicine 151, 562–569. Make, B.J., Hill, N.S., Goldberg, A.I., Bach, J.R., Criner, G.J., Dunne, P.E., Gilmartin, M.E., Heffner, J.E., Kacmarek, R., Keens, T.G., 1998. Mechanical ventilation beyond the intensive care unit. Report of a consensus conference of the American College of Chest Physicians. Chest 113, 289S. Man, J.P., Sin, D.D., Ignaszewski, A., Man, S.F., 2012. The complex relationship between ischemic heart disease and COPD exacerbations. Chest 141, 837–838. Quanjer, P.H., Borsboom, G., Brunekreef, B., Zach, M., Forche, G., Cotes, J., Sanchis, J., Paoletti, P., 1995. Spirometric reference values for white European children and adolescents: Polgar revisited. Pediatric Pulmonology 19, 135–142. Sassoon, C.S., Light, R.W., Lodia, R., Sieck, G.C., Mahutte, C.K., 1991. Pressure–time product during continuous positive airway pressure, pressure support ventilation, and T-piece during weaning from mechanical ventilation. American Review of Respiratory Disease 143, 469–475. Tobin, M.J., Laghi, F., 1998. Monitoring of respiratory muscle function. In: Tobin, M.J. (Ed.), Principles and Practice of Intensive Care Monitoring. McGraw-hill Companies, New York, pp. 497–552. Tuxen, D.V., Lane, S., 1987. The effects of ventilatory pattern on hyperinflation, airway pressures, and circulation in mechanical ventilation of patients with severe air-flow obstruction. American Review of Respiratory Disease 136, 872–879.