Optimized algorithm for speed-of-sound-based infant sulfur hexafluoride multiple-breath washout measurements

Abstract

Introduction: Major methodological issues with the existing algorithm (WBreath) used for the analysis of speed-of-sound-based infant sulfur hexafluoride (SF₆) multiple-breath washout (MBW) measurements lead to implausible results and complicate the comparison between different age groups and centers.

Methods: We developed OASIS-a novel algorithm to analyze speed-of-sound-based infant SF₆ MBW measurements. This algorithm uses known context of the measurements to replace the dependence of WBreath on model input parameters. We validated the functional residual capacity (FRC) measurement accuracy of this new algorithm in vitro, and investigated its use in existing infant MBW data sets from different infant cohorts from Switzerland and South Africa.

Results: In vitro, OASIS managed to outperform WBreath at FRC measurement accuracy, lowering mean (SD) absolute error from 5.1 (3.2) % to 2.1 (1.6) % across volumes relevant for the infant age range, in variable temperature, respiratory rate, tidal volume and ventilation inhomogeneity conditions. We showed that changes in the input parameters to WBreath had a major impact on MBW results, a methodological drawback which does not exist in the new algorithm. OASIS produced more plausible results than WBreath in longitudinal tracking of lung clearance index (LCI), provided improved measurement stability in LCI over time, and improved comparability between centers.

Discussion: This new algorithm represents a meaningful advance in obtaining results from a legacy system of lung function measurement by allowing a single method to analyze measurements from different age groups and centers.

Keywords: OASIS; WBreath; lung clearance index.

Figure 1

Illustration of the signal processing steps of OASIS. (A) Molar mass (MM) signals during the pre‐phase and end of washin as a function of inspired/expired volume of an example measurement (black lines). From those, a median trace is extracted for each of the four boundary conditions above (expirations/inspirations at minimum/maximum sulfur hexafluoride [SF₆]). (B–D) Shown is the signal of the first 2 s of the washout phase of an example measurement, showing a complete breathing cycle (inspiration & expiration). (B) Tidal change correction (TCC): Shown is the raw molar mass signal (black), expirogram functions for the end of washin and pre‐phase, and the corresponding inspirogram functions from panel (A). The dashed respirogram functions correspond to the maximum signal (4% SF₆), the solid lines to the minimum (0% SF₆). (C) Side chamber correction: Shown is the output of the TCC in panel B) (black). The minimum signal is fit separately to each inspiration as a function of inspired volume, yielding the signal corresponding to 0% SF₆ during inspirations (solid blue). The discontinuous blue curves are interpolated to yield the curves corresponding to 0% SF₆ during expirations (solid red). The dashed curves (corresponding to maximum signal) are scaled versions of the solid curves, scaled to ensure that they correctly yield known solutions for the first breath of washin and washout. (D) Final extracted SF₆ signal (black).

Figure 2

In vitro functional residual capacity (FRC) measurement comparison between WBreath (config II, black markers), and OASIS (empty markers). Shown is the absolute error from the target volume of the in vitro lung model. Data points represent the mean of two triplicates each (initial measurement and repeat on a device calibration). Standard: baseline condition (T = 32.5 ± 1°C, small: FRC = 80 mL, VT = 30 mL, RR = 30/min, large: FRC = 210 mL, VT = 50 mL, RR = 20/min), −T: measurement performed at room temperature, +RR: increased respiratory rate (small: +66%/large: +50%), +VT: increased tidal volume size (small: +66%/large: + 60%), +VI: added ventilation inhomogeneity mesh within lung model.

Figure 3

Comparison of infant data using WBreath (config I), WBreath (config II), and OASIS. Shown are the primary outcomes (FRC, LCI) of MBW measurements in n = 48 healthy control infants measured at age 8 weeks (8w) and 1 year (1 y) from the Drakenstein cohort (South Africa), analyzed using three different methods. New: Measurements analyzed using OASIS, config I: WBreath with config I settings (Drakenstein cohort settings), config II: WBreath with config II settings (BILD/SCILD cohort settings). Boxes represent median + interquartile ranges. (A) LCI in turnovers (TO) analyzed with different methods, showing the change for each visit between the different methods. (B) FRC in (mL/kg) analyzed with different methods, showing the change for each visit between the different methods. Statistics: Paired t‐test: ***p < .001, **p < .01, *p < .05, ns, not significant. BILD, Basel‐Bern Infant Lung Development; FRC, functional residual capacity; LCI, lung clearance index; MBW, multiple‐breath washout; SCILD, Swiss Cystic Fibrosis Infant Lung Development.

Figure 4

Longitudinal change in lung clearance index in children with cystic fibrosis. Shown is lung clearance index (LCI) in turnovers (TO) for n = 21 children from the Basel‐Bern Infant Lung Development (BILD) cohort (Bern) measured at 8 weeks (8w, infant sulfur hexafluoride [SF₆] multiple‐breath washout [MBW], either WBreath or OASIS), 1 year (1 y, infant SF₆ MBW, either WBreath or OASIS), and at 6 years (6 y, N₂ MBW, Exhalyzer D, Spiroware 3.3.1). Boxes show median + interquartile ranges. (A) Longitudinal analysis using WBreath for infant MBW measurements. (B) Longitudinal analysis using OASIS for infant MBW measurements. Statistics: Paired t‐test, ***p < .001, *p < .05, ns, not significant.

Figure 5

Multiple‐breath washout (MBW) outcomes in healthy control 6‐week‐old infants over time and across centers. (A and B) Shown is lung clearance index (LCI) in turnovers (TO) for n = 78 children from the Basel‐Bern Infant Lung Development (BILD) cohort (Basel) measured at 6 weeks, as a function of test date. Black dots indicate measurements before January 2017 (n = 53), empty circles measurements after January 2017. (A) Analysis using WBreath. (B) Analysis using OASIS. (C and D) Shown is LCI in TO for infants in South Africa (n = 48, Drakenstein cohort, 8 weeks old), Bern (n = 62, BILD cohort, 6 weeks old) and Basel (n = 78, BILD cohort, 6 weeks old). (C) Center comparison using WBreath (SA: config I, Bern: config II, Basel: config II). (D) Center comparison using OASIS. Boxes show median + interquartile ranges of the two sub‐groups. Statistics: Unpaired t‐test ***p < .001, **p < .01, *p < .05, ns, not significant.